A new MuseRig account opens with 1000 credits. You do not need a card to get them and there is nothing to cancel.
What 1000 credits is worth
One of each of the three things most people come here to make:
549 credits — a product ad on the Quality finish, 15 seconds.
349 credits — a presenter video on the Quality finish, 30 seconds long.
79 credits — a product photo set on the Quality finish, which is four images.
That is 977 of the 1000. You do not have to spend it that way, and the 23 left over will not buy a fourth piece. The point of the three is that you see what each recipe does on your own work before you decide which one is worth paying for.
They are ordinary credits
Free credits buy the same outputs at the same prices as credits you pay for. There is no cheaper free version of a photo set and no watermark on a free presenter video. You still see the exact price before each piece runs, and you still approve it yourself.
That is deliberate. A free sample made by a cheaper process would tell you nothing about what you would get after paying.
What they will not stretch to
A product ad on the Premium finish is 1149 credits, which is more than the grant. A Faceless Video at 599 or 899 is reachable, but not alongside the three pieces above.
The grant is sized to show you the product on your own work, not to finish a campaign. The fourth piece is the one MuseRig asks you to pay for.
They do not refill and they do not expire
There is one grant per account. It does not come back next month, so there is no monthly rhythm to plan around and no reason to hurry — unspent credits sit there until you use them.
Watch the 30-day rule while you are still deciding. Any output you do not favorite is deleted 30 days after it is made, free or paid. If you like something, favorite it before you close the tab.
When they run out
The balance stops you at the quote, not halfway through a generation. You see what the piece costs, what you have left and how far short you are, and nothing runs until you top up.
Everything already made stays yours. Buying credits does not change any price you have already been shown — the posted price is the posted price whichever way you paid for it.
Every early-access account has a link to share. Somebody who signs up through it takes one of the remaining seats, and once they have made their first piece you get 500 credits.
Where the link is
Open Settings and find Bring someone in. Your link is there with a Copy button. It is also one tap from the credit meter in the sidebar.
What the person you invite gets
A seat, while seats last. Early access is a fixed number of accounts, first come first served, and the number left is printed beside your link. They get the same free credits every new account gets, and no card is asked for.
When you are paid
Not when they sign up. When they finish their first piece: a photo set, an ad, a film, anything MuseRig makes. Signing up proves an inbox; finishing something proves a person. The credits land in your balance the moment their piece does, and MuseRig sends you one line of email so you do not have to check.
You are paid for the first two people who do this. The link keeps working after that for anyone who wants a seat; it just stops paying you.
What you can see
Three counts under your link: how many signed up, how many made something, how many credits you have earned. Not who. The people who use your link have not agreed to hand you a list, and the founder's view is the only place their names appear.
Sometimes a line says one is waiting on review. That means something about the two accounts looked alike enough for a person to check before paying, and the founder will. It is not an accusation, and it is often a household.
What does not count
Two accounts reaching the same inbox are one person, however the address is spelled, and the second one is not opened. A member cannot be paid for themselves. And nothing is paid for an account that never makes anything, however long it sits.
MuseRig works best when you begin with the business outcome, not a technical prompt. Choose a recipe, add the references you already have, and describe what the finished piece needs to achieve.
1. Choose the closest recipe
MuseRig includes product photography, a 15-second product ad, a 30-second faceless explainer, a reusable spokesperson, and a picture film made of photographs you already own. Each recipe asks only for decisions that matter to that kind of work.
The picture film is the odd one out and worth knowing about before you choose: it generates nothing at all. Your photographs are laid on a plain wall with a gallery caption beneath each one, held, and dissolved between. Everything else here asks a model for pixels.
2. Review the visible plan
Before anything costs money, MuseRig keeps the references, brief, quality, shape, planned outputs, preservation rule, and exact total visible for review. The cost remains at zero until approval.
For a product photo set or a product ad, the shape is yours to choose, named by where the piece will go: feed portrait for Instagram and Facebook, square for a Google listing or LinkedIn, vertical for Shorts and Reels, landscape for YouTube and a website. A photo set offers all four; an ad offers the last three, because its route cannot make 4:5. Left alone, a photo set is feed portrait and an ad is vertical, which is most common and recommended.
3. Approve the exact quote
The quote shows dollars, credits, planned outputs, and failure protection. A generation begins only after you approve that exact quote.
Nothing is spent while you plan. Failed generation work is refunded automatically.
4. Keep the Library useful
Use the Project shelf filters to separate drafts, completed work, and failed attempts. Archive hides a project from the active Library without changing its status or deleting its media, job history, quotes, or accounting; the signed-in human confirms the move and can restore it later. A failed project can be retried as a new draft, leaving the original failure record intact.
Connected agents can read active projects, prepare reversible drafts, and archive, restore or move a project to trash, telling you afterwards. All three are reversible from the Library. Deleting an image or a saved presenter is not reversible, so an agent can only ask and you decide.
5. Keep what you want to keep
MuseRig does not keep every draft forever, and it should not surprise you when it clears one out.
Kept until you delete them: favorited outputs, your reference images, brand assets, private music and sound effects, and saved characters.
Deleted after 30 days: any generated output you neither favorited nor downloaded. Clicking the heart again takes the favorite off, and the output goes back on the 30-day clock unless you have downloaded it. Downloading keeps an output the same way favoriting does; the favorite is the one you chose, and downloading does not move it. The clock starts when the output is made, and the deletion date is set at that moment rather than applied retroactively.
When a set finishes, Compare on the Output Board puts every direction side by side at the same size, whole, with the same favorite and download controls, so you choose from one screen. When you like one of the directions, favorite it. That single click promotes the output to permanent and clears its deletion date; the others go back on the 30-day clock. Nothing about your project, quote, job history, or accounting record is removed either way — those stay whatever happens to the media.
Archiving a project is not the same thing. Archive only hides a project from the active Library; it deletes nothing and starts no clock.
A blank brief is the hardest part of a first project. So the Library carries five worked examples, one for each recipe, with the brief, the plan and the shape already filled in. Read one, then make it yours. None of this costs a credit.
What they are
All five are about SUMMIT/01, a fictional electrolyte drink. SUMMIT/01 is a made-up product and brand. It is not a real business. The website on its end cards, summit01.example, is a reserved address that belongs to nobody, so nothing in an example can be mistaken for a real business.
Launch photo set - a product photo set at Quality, feed portrait. The brief says what to keep exact and what to leave out before four images are made from your own product photos.
15-second launch ad - a product ad at Quality, social vertical. The brief for the footage, then the three places words go: a hook, an offer and an end card. It also says what to do with the sound that comes back: lowered under the captions.
30-second explainer - a faceless video at Quality, landscape. This is the plan behind MuseRig's own explainer film: a 59-word script, six five-second scenes with their captions, narration direction and a sound choice.
20-second presenter - a synthetic spokesperson at Quality, social vertical. A 38-word script sized to a 20-second window, a voice, a film shape, a disclosure choice and an end card. The presenter is the one thing it does not choose.
30-second picture film - photographs you already own, laid out and held, with nothing generated. The example sets the length rather than the pace: say thirty seconds and the holds are worked out from it, so adding a picture shortens each one instead of making the film longer than the feed it was made for. It arrives with no pictures at all, because this is the one recipe that invents nothing and yours are the whole of it.
Reading one
Open the Library and expand Examples. Each card shows what the example is for; Read the plan shows the brief and the plan in full, the way the workbench will show them once the plan is yours. Nothing on the shelf can be edited or quoted.
Making it yours
Press Make this mine. MuseRig creates a draft in your workspace with every field the example carries, and opens it in the workbench. From then on it is an ordinary draft: rename it, change anything, save it, quote it.
Four things do not come with it, because they are yours to say:
Reference images. The example brings none. A photo set and an ad are made from your product photos; the explainer and the presenter video build their end card from one; and a picture film is made of nothing else.
A presenter. The presenter example names none. Choose one from your Characters.
Rights and consent. The synthetic rights confirmation and the consent review are statements you make about your own work. No example can make them for you.
Music. An example never chooses a track. If you want music, choose it from your Audio shelf and confirm its licence for this project.
Each card also says what to change first, in the order you will meet it.
From a connected agent
list_example_projects reads the same five examples, whole, under the read permission. create_project_draft with an exampleSlug makes one the owner's draft. It takes only a name beside the slug; anything else is changed afterwards with update_project_draft, where the change can be seen. Neither spends a credit, and the draft arrives with the same four things missing.
Making an example yours spends nothing. The first credit moves when you approve a quote, and not before.
Use one to four clear images in ordinary light. Include the front, the overall shape, and any label or packaging detail that must remain exact.
A useful four-image set
A straight-on front image for the label and logo.
A three-quarter image for shape and depth.
A side or back image if packaging geometry matters.
One close detail showing important materials or texture.
Avoid filters, motion blur, extreme wide-angle distortion, and tiny screenshots. A plain phone photo with honest detail is usually more useful than a heavily styled image.
MuseRig accepts JPEG, PNG, and WebP files up to 10 MB each, in any shape between half as wide as it is tall and twice as wide; a taller or wider picture is refused when you add it, with its size in the message, because the model would refuse it later after your credits were reserved. Crop it closer to the shape of the finished piece. Product references stay private and remain in your workspace until you delete them. Generated drafts you neither favorite nor download—not your source product references—are eligible for cleanup after 30 days.
The Image shelf
Every image you add belongs to the workspace, not to one project. The Image shelf in your Library shows all of them with their filenames, and anything on it can be used as a reference, an end card or a cutaway in any project without uploading it again. You can add images from the shelf or from the step that needs one.
Making an image
Three ways to get a picture you do not have. A Product photo project needs no reference: describe it, see the price, approve it, and MuseRig returns four to choose from. Help me create an image on the Image shelf writes that brief for you from three answers, and writing it costs nothing; then either make it in MuseRig for the posted price, or copy it as a prompt for ChatGPT or another image tool and add the picture you get back to the shelf. If you have a connected agent that has passed the read-only check, a third button, Copy a prompt for your agent, hands it the brief: an agent with an image tool of its own, such as Codex, makes the picture and puts it on your shelf; one without says so and stops. Either way the picture arrives marked as added by your agent and waits for you to open it. Never ask for words inside a picture, because generated lettering comes out wrong; MuseRig lays exact wording over the top on end cards, captions and text cards. A photo set built from a description alone is now told to draw no lettering at all, so a made-up sign or menu should not appear in one. A set built from your own product photographs is told the opposite, to keep the packaging, label and logo exactly as they are, because those words are the point.
Keep on shelf on the Output Board keeps a generated image on your Image shelf, so other projects can use it. It saves a copy, so the output still follows its own 30-day rule. The copy arrives as a style reference: something to work from, rather than a photograph of your real product.
Images your agent adds
A connected agent can add an image to the shelf for you. It needs to be one that can send a file, such as Claude Code, Codex or Cursor; an assistant that only chats cannot. Some agents can also make the picture: Codex carries an image tool of its own, so it can draw from a brief and add what it drew. MuseRig never asks an agent to draw and cannot tell whether yours can; it tells every agent how a picture it makes reaches the shelf, and the agent, which knows its own tools, does it or says it cannot. An agent that cannot draw can still write the brief, and you make it here or in your own image tool. The paid Product photo route is unchanged either way.
An image your agent adds arrives marked added by your agent, not opened yet. It is on the shelf and usable straight away, but a quote for any project that uses it is held until you have looked at it. Click it on the shelf, see what it is, close it, and the quote goes through. Your agent can be wrong about a picture in ways it cannot see, and a render costs credits, so the last look is yours.
An agent cannot add a presenter likeness. Those are yours to add on the Characters page.
The shelf shows as a Grid of tiles or as Cards, one row per image with its size, type, when it was added and who added it; the choice is remembered in this browser. Removing an image takes it out of every picker. MuseRig refuses if a character uses it as their likeness — remove that character on the Characters page first — or if a project you have already paid to generate names it anywhere, as a reference, an end card or a cutaway, so finished work keeps its provenance. A project you only asked the price of does not hold an image.
MuseRig records one image-rights confirmation, and only one: the synthetic-presenter statement at the Spokesperson Consent stage, which covers that presenter's reference. It records nothing for a product shot, a logo, an end-card image or a picture in a picture film. Having the right to use those is yours to know, and reusing one from the shelf does not change that either way.
That last one is worth saying plainly rather than leaving in a list. A picture film can be made entirely of somebody else's work, and MuseRig asks you nothing about it. If the paintings are not yours, you need the artist's permission to make the film and to publish it. Nothing here checks, and no step going through without a warning means anything has been cleared. The Terms say the same in section 6.
For printed words and logos, a clean front-facing reference is more valuable than a beautiful lifestyle shot.
Open Connections in MuseRig and choose the app you use. MuseRig shows the exact place to click and the one secure address to paste.
Setting it up
In MuseRig, open Connections and choose ChatGPT, Claude, Codex, Cursor, or Other.
Follow the exact first step shown for that app. MuseRig also tells you when a paid plan, workspace permission, or local command is required.
Copy the setup value MuseRig shows. Most apps use https://muserig.com/mcp as the server address; Codex uses the two displayed terminal commands. You never need a provider API key or a hand-written configuration file.
When your app opens MuseRig, sign in and review the requested permissions.
Return to Connections and copy the harmless read-only test prompt. Three status boxes show whether the MuseRig server, authorization, and tool check are working.
Safe connection test
After authorization, paste this exact prompt into the agent you connected:
Use MuseRig's muserig_connection_check tool only. Report the account and granted permissions. Confirm from the tool result that this check spends no credits and changes no workspace data. If muserig:generate-request is granted, explain that it can only prepare a quote and request human approval; it cannot spend credits or start generation. Do not create or update a draft, request a quote, or request generation.
The response should report your account and permissions and explicitly show spendsCredits: false and changesWorkspace: false. No other MuseRig tool is needed for this test.
If muserig:generate-request appears in the granted permissions, it means the agent can prepare a quote and return a request for your approval. That permission cannot spend credits or start generation by itself.
Start with Read + Draft. Your agent can read your brand kits, correct one, add one, and prepare projects, none of which spends credits. Paid generation returns to MuseRig for approval unless you later add a tightly limited spending allowance.
Your agent never needs your MuseRig password or a generation-provider API key.
If the connection fails
Confirm that the address begins with https:// and ends in /mcp, that you signed in with your MuseRig email, and that your agent supports remote MCP servers that open a browser for authorization. The Connections screen reports compatibility problems in plain language.
If Create, Add custom connector, or MCP is missing from your agent app, the feature may not be included in your plan or may be disabled by a workspace administrator. That is an agent-app limitation, not a MuseRig password or storage problem.
For the Codex CLI, MuseRig provides two copyable commands: one adds the remote Streamable HTTP server and one opens the OAuth sign-in. Do not create a JSON file or paste a provider key.
What is in the next chapter
What a connected agent does on its own, the two things it can only ask you for, and the four more things Draft access covers are all in What a connected agent can do.
Three things an agent does and tells you it did, two it can only ask for, and four more that Draft access covers. Getting connected, the read-only test that proves it, and what to do when the connection fails are in Connect ChatGPT, Claude, Codex or Cursor.
An agent asks, you decide
An agent can ask to delete your work. Only you can delete it.
Two things a connected agent can only ask about, because nothing can undo them:
Delete an image from the shelf — the upload is gone. Drafts using it as a reference, an end card or a cutaway will need another image.
Delete a saved presenter — the presenter and their likeness leave your workspace. Finished films already made with them are untouched.
The request names the exact image or presenter, and carries the agent's reason in its own words. "Delete this image" and "delete this image, nothing references it any more" are different asks, and you should get to read the difference.
Requests appear in your Library. Sign in, read the row, approve or decline. Approving is what does the work; declining ends it. Neither of these can be undone, and the row says so before you press anything.
Three it does without asking
Archiving a project, restoring one and moving one to trash are things an agent simply does, and tells you it did.
Nothing is lost by any of the three. Everything stays where it was and restoring is one step, so asking your permission each time would only teach you to approve without reading. The two above are the ones worth arriving at still paying attention.
An agent will not touch a project that is mid-generation, or one whose quote you approved. Tidying up around a running job is how somebody loses work they are waiting on. A quote nobody approved is only a price, so a project sitting on one can still be archived.
A request is worth deciding for 24 hours. After that it expires and the agent has to ask again, because an agent asked about your workspace as it was when it looked, and a day later nobody has checked.
Your agent can also read back what it has asked for and what you decided, so it does not ask twice for the same thing.
One rule behind all of it
Nothing an agent can reach through MuseRig destroys anything. The tool it uses to ask removes nothing by itself: it writes down a request, and the work happens only when you approve it.
Four more things Draft access covers
All four sit inside the same rule: an agent may choose from a shelf you vetted, and may never make a statement on your behalf.
Choosing and saving a presenter. If you have presenters saved and ready to generate, an agent can bind one to a Spokesperson project, the same way it can name one of your images for an end card. It can also save a new synthetic presenter — a name, one of the eight curated voices, and whether they are illustrated or photoreal — because none of that says anything about a real person.
What it cannot do is give them a face or make them filmable. Attaching a likeness, building the Runway avatar and confirming the rights to a reference are all yours, and so is approving what it costs to film anyone. A presenter an agent saved sits there with no portrait until you add one.
Adding an image to your shelf. An agent can put an image where your projects can use it, in two steps: it asks MuseRig for a private upload link, then sends the file straight to storage. The bytes never pass through the agent's conversation, and MuseRig never fetches an address the agent chose. The image lands marked added by your agent, not opened yet, and a quote for any project naming it is refused until you have opened it on the Image shelf. It cannot add a presenter likeness, replace or delete an image, or confirm the rights to one.
Your agent needs to be able to send a file for this. Claude Code, Codex and Cursor can; an assistant that only chats cannot. An agent with an image tool of its own, which Codex has, can also make the picture: MuseRig tells every agent how a picture it makes reaches the shelf, and an agent that can draw does it while one that cannot says so. MuseRig never asks an agent to draw and cannot see which tools it holds. An agent that cannot draw can still write the brief, which you then make on the Product photo route or in your own image tool.
Reading what a plan will actually do, before anyone pays. For a Spokesperson project an agent can ask how long the film is likely to run — a fastest and a slowest figure for the exact words in the script — and, for each cutaway, whether it renders, renders only sometimes, or never renders. That is the same check the workbench shows you, and an agent that can read it can move a cutaway that would otherwise have been dropped after you paid for the render.
The same question works on the other routes. On a Faceless Video it reads the end card and which stages you have marked; on a picture film it reads the running order back as it will play, with every picture's title, the hold each one gets and the total length, so an agent can answer how long the film is without guessing.
Copying a project you already reviewed. An agent can duplicate a project, which carries the plan, the captions and the running order into a fresh draft and costs nothing. Two things deliberately do not come with it: the copy needs its synthetic rights confirmed again, and its music chosen again. Both are statements you make, and a copy is not a reason to inherit one.
Your website already says what you sell, who it is for, what colours you use and how you talk. A connected agent can read it and fill in MuseRig for you - the Brand kit and a first draft - and then stop, so you read and correct it before a credit moves.
What happens
The agent reads your site.
It looks at the brands you already have. If one for this site is saved, it corrects that kit. If none is, it adds one. Either way it writes the business name as the site writes it, the colours the site actually uses, the truths the site states in its own words, and two sentences on how the site speaks.
It drafts one product photo set of your main product, with a brief written from the site.
It stops. No quote is requested and nothing is generated. The agent cannot spend credits whatever it is told; MuseRig's server refuses, not the agent's good manners.
Every line the agent writes to the kit is shown to you in full. Fix anything the site got wrong, or anything the agent read wrong, on the Brand kit page.
A workspace holds a set number of brands, five to begin with; the Brand kit page shows the count. Correcting one and starting another are different acts, and the agent has a separate tool for each, so an agent asked to fix a colour cannot end up creating a second business by accident.
The prompt
Open the Brand kit page, type your website into Start from your website, and copy the prompt it gives you. Paste it into the agent you connected. With no website typed in, the prompt reads:
Read https://yourshop.example and use MuseRig's tools only.
1. Call list_brand_kits to see which brands are already saved.
2. If a kit for this site is already there, call update_brand_kit with its brandKitId. If none is, call create_brand_kit. Either way send: name - the business name as the site writes it; colors - the two or three colours the site actually uses, as hex codes where you can read them; truths - only facts the site states about the business or its products, one per line, in the site's own words, never a claim you inferred; voiceNotes - two sentences on how the site speaks to its customers.
3. Call create_project_draft for one product photo set of the main product, with a brief written from the site, and stop there.
Do not request a quote and do not request generation. Report everything you wrote to the kit, in full, so the owner can correct it before anything is made.
What the agent may not do
It may not invent a claim. The prompt says so, and a truth is only worth having if the site actually states it. It may not confirm rights or start anything paid, and an image it adds to your shelf waits for you to open it. It writes drafts, and drafts are reversible.
The Brand kit lines the agent writes are the lines every later photo and ad in that brand carries. Read them once, properly, and every project in that brand inherits what you checked.
One caveat on length. The models take a prompt of a fixed size, and your brief and MuseRig's own direction are kept whole, so a kit with a great deal to say may have its last lines left off a particular run to fit: the voice notes first, then the truths, then the colours. Nothing you wrote is lost or changed, and the brand name and colours are the last to go. If your kit is long and you want a particular truth to reach the model every time, say it in the project's brief as well.
Every generation plan includes an exact retail quote and a credit amount. The quote expires after fifteen minutes so a stale provider price can never be approved accidentally.
Which finish to choose
Quick is for comparing directions. Run two or three briefs at Quick, keep the one that works, then run it again at the finish you will publish.
Quality is the finish for almost everything you will post. It is the tier MuseRig has measured most, and the free credits are sized around it.
Premium is for when the product being wrong is the failure, not when you want sharper pixels.
On a product ad the difference is not resolution. Quick and Quality open on your reference photo and let the model move; the bottle or the cup can drift by the closing seconds. Premium runs a different recipe that holds the product exact in every shot, label and all. Choose it when a label must read correctly, when packaging will be compared with the real thing, or when the ad will run paid and a second run to fix drift would cost more than the tier. A drink of the week for your own feed does not need it. On a product photo set, Premium asks the image model for its highest-effort render of each picture. On a picture film it is 4K for a gallery screen where Quality is 1080p. A presenter film has one finish, Quality, because the presenter is made the same way whatever it is called.
Faceless video has no Premium, and the presenter film has neither Quick nor Premium: in both, the other tiers produced what Quality produces, so they were removed rather than sold at a difference nobody could see.
A recipe only shows the tiers it actually sells. A picture film has no Quick, for the opposite reason to the one above: its Quality is 1080p and its Premium is 4K, which is a real difference, and a 720p film of somebody's painting is not a thing to sell an artist when the whole promise is that the work arrives intact.
Reroll only what needs work
On a finished product photo set, favorite the image that did not work and press Remake this one. MuseRig quotes that one image at the posted per-image price - 10, 20 or 40 credits for Quick, Quality or Premium - and the other three stay exactly as they were. The new image takes the old one's place in the set once you approve the quote.
Remaking one scene of a faceless video or one moment of a spokesperson film is not built yet; for those, favorite the strongest output and create a fresh project if you need another pass.
MuseRig never starts another generation silently. Failed provider work releases the reservation and restores the credits.
Credits are what MuseRig charges in. One credit is one cent of generation value, so a quote
of 79 credits is a piece of work that costs 79 cents.
What a pack costs
Two packs, both at face value:
1,000 credits for $10
5,000 credits for $50
A pack grants exactly what it costs. There is no volume discount, so the larger pack is the
same deal bought in one go rather than a better one — there is nothing to work out and no
reason to buy more than you need.
Nothing renews and nothing expires. A pack is a one-time purchase, there is no subscription
behind it, and buying once does not sign you up to buying again. Unspent credits stay on your
balance until you use them, whether that is next week or next year.
How to buy
Add credits, under the credit meter in the sidebar. It shows both packs and what each one
is worth in finished work.
Payment is handled by Stripe. Your card details go to Stripe directly, and MuseRig never
receives or stores them. The charge appears on your statement as MUSERIG, and the seller
is Blueridge Web Services LLC, the company that operates MuseRig.
What they do
Bought credits behave exactly like any other credit. They buy the same outputs at the same
posted prices. You still see the exact cost of a piece before it runs, and you still approve
it yourself.
Provider work that fails releases the credits it reserved, automatically. Credits are not
otherwise refundable, so buy for the work in front of you rather than stocking up.
A full balance is not permission to spend it. Nothing is charged until you approve an exact
quote, whichever way the credits arrived.
A 15-second product ad is footage from the model plus words from you. The words are laid over and after the footage by MuseRig, in your exact spelling, at a size that reads on a phone. Nothing you type is handed to the model to draw.
Why the words are yours
Video models still get text wrong. A price comes out as a shape that looks like a price; a website comes out with a letter missing. So MuseRig does with the ad what it has always done with the faceless explainer: the model makes the moving picture, and the exact words go on afterwards from what you typed. Every product ad prompt also tells the model to draw no lettering of its own: no signs, menus, chalkboards, prices or captions in the footage, only what a product's own label already shows. If your brief asks for words on screen, put them in the hook, the offer or the card instead.
Three places for words
The hook. Up to 48 characters over the opening four seconds. The one line that stops the scroll: what it is, or who it is for.
The offer. Up to 48 characters over the closing four seconds, before the card. The price, the deal, or what to do next.
The end card. Two seconds at the end, on MuseRig's paper, on a colour of your own — your Brand kit's hex colours are offered first, and the type goes pale on a dark ground — or over one of your own reference images. Brand name, tagline, website and a closing line. It is the same card a synthetic spokesperson film ends on, so what you learn here holds there. The card is drawn under its fields at the shape of the ad, and is edited on that picture: drag a line or the logo to move it, double-click a line to type into it where it sits, or select it and use the arrow keys, and set its size and colour beside the card. Your browser draws that picture so it keeps up with you; See the film's own render shows the one MuseRig draws for the ad itself. Both come from the same measurements, so what you place is what plays.
Leave any of the three empty and it is not rendered. There is no blank caption and no empty line on the card.
The card has to be marked reviewed before the quote: press Mark the end card reviewed under it once it reads right, and any later change to the card asks for the mark again. The quote refuses until it is there, and a connected agent cannot make it.
The closing line
The card's last line carries the MuseRig credit until you change it. Changing or removing it is available once you have bought credits; free-allowance credits are a grant, not a purchase. Put an email address or a phone number there and it is set in contact weight, large enough to type back from a phone.
The sound
The ad asks the model for sound as well as pictures. What comes back varies: sometimes music, sometimes the sound of the scene, sometimes almost nothing. MuseRig measures the file before the words go on and names it - silent, speech, music, ambient sound, or speech over music - and the Output Board says which once the ad is made: "Sound: music, kept."
You decide what happens to it, on the same step as the words:
Keep the sound. Whatever arrived plays as it is, brought up to the same level as every other MuseRig film.
Lower it under the captions. The sound drops by 10 dB while the hook and the offer are on screen, and comes back between them.
Mute it. The ad plays silent.
Replace it with my music. One Music track from your Library, with a rights confirmation for this project. The model's sound is dropped and your track carries the whole ad. The quote waits until you have chosen the track.
The measurement is a measurement, not a listen. A voice over loud music reads as music, and a steady wind can read as music too. If the board's word surprises you, play the ad.
What it costs
The finished ad is 17 seconds: 15 of footage and 2 of card. The exact layer is rendered in MuseRig's own compositor and is already inside the posted price. The quote does not change because you typed a hook.
Before you approve
The quote panel reads the plan back beside the price: the hook, the offer, the card, the closing line, the sound and its track, the shape, and the reference. Those lines are what renders. Read them there, on the Review step, before the button.
Check the spelling before you approve the quote. What you type is what is rendered - MuseRig corrects nothing, and that is the point.
Faceless Video uses six five-second scenes plus a two-second end card. Planning is deliberately separated from generation so you can inspect the entire argument before MuseRig asks for a paid approval.
The seven review stages
Source and destination — State the business outcome, audience and trustworthy source. Choose Social vertical, YouTube, Square feed or Feed portrait, then add at least one owned visual reference. MuseRig uses it for exact product proof and the deterministic final end card.
Script — Review the complete 30-second argument. The pace estimate helps you catch a script that is too thin or too crowded.
Six scenes — Edit the visual direction, narration and exact on-screen caption for every five-second scene. The narration is measured as you type. The six narrations are read as one voice track over the scenes, with room for about 61 words in all at the pace MuseRig has measured; a longer one loses its last words, so the quote refuses until it is shortened, and says by how much. About 10 words keep the voice on the scene it describes.
Captions and sound — Choose a readable caption treatment, then choose Narration only, Narration plus my licensed music, or Intentional silence and write a reversible sound brief. No voice or media is generated during this stage. The signed-in human can choose one reusable Library music bed and up to three Library Sound FX placed on exact scene boundaries, or upload a private licensed music track, before reviewing the stage.
End card — Review the closing two seconds as they will render: brand name, tagline, website and closing line on MuseRig's paper, on a colour of your own, or with one of your own images. It is the same card a synthetic spokesperson film and a product ad end on. Drag any line or the logo on the picture to move it, double-click a line to type into it where it sits, or select it and use the arrow keys, and set its size and colour beside the card. Your browser draws that picture so it keeps up with you, and See the film's own render shows the one the film plays; both come from the same measurements. A new draft under a Brand kit starts with the kit's name and logo. Any change to the card reopens this stage and the quote after it.
Quote — Review the quality route, assembly plan and current exact plan price. This still spends nothing.
Explicit approval — Only the signed-in human can approve the rechecked quote and allow credits to be reserved.
Generated footage can supply movement and atmosphere. MuseRig keeps exact interface text, prices, captions, logos and domains in the deterministic composition layer rather than trusting a video model to reproduce them.
After approval, MuseRig saves the job before it starts and shows a stage-weighted estimated progress range across provider generation, narration, composition, verification and Library persistence. You can safely leave the page and reopen the project later. Time remaining is learned only from comparable successful jobs; until enough evidence exists—or when a stage takes longer than the learned range—the app says it is learning or recalculating rather than presenting false precision. A job reaches 100% only after the output has been verified and saved to the Library.
The finished film is 32 seconds: 30 seconds of scenes plus the 2-second end card. Quick plans one silent 15-second generated sequence for the kinetic middle of the film, while the opening, owned-product proof and end card stay exact. Quality plans two ordered generated sequences at the higher resolution. Narration is generated separately, normalized and mixed during the isolated FFmpeg composition step so generated footage cannot silently change the approved words.
The current Sound and narration choice exposes only what the live route can deliver honestly: Narration only uses the reviewed voice track without added music; Narration plus my licensed music mixes one private or reusable rights-confirmed track beneath narration; Intentional silence creates no narration, music or Sound FX. Up to three selected Sound FX begin at their chosen five-second scene boundaries and are mixed during deterministic final composition. A connected agent can revise the mode and brief, but cannot select audio, place Sound FX, or confirm rights. Any sound change reopens the human sound review and quote review.
The Library Audio shelf retains reusable music and sound effects with their source and license records. Saving an item never silently attaches or approves it. For each project, the signed-in human must re-confirm that the current license permits the intended use before choosing the music or Sound FX. A connected agent can research candidates and save a metadata-only shortlist for you to look at, with sources, licence links and prices. It cannot buy, download, select or attest to any of them.
YouTube is a first-class 16:9 destination. Social vertical uses 9:16 for Shorts, Reels and TikTok; Square feed uses 1:1; Feed portrait uses 4:5. A destination change is reversible draft work, clears the prior review trail and spends nothing.
A connected MCP agent can prepare or revise every reversible draft field. It may request the approval screen, but it cannot approve spending or start generation.
A picture film is the one MuseRig recipe that generates nothing. It takes photographs you already own, lays each on a plain wall with its caption beneath, holds it long enough to be looked at, and dissolves to the next. MuseRig supplies the typography and the timing. The pictures are yours and they arrive on screen exactly as you sent them.
That is worth being blunt about, because every other recipe here works the other way round. No model is asked for anything, nothing is invented, nothing is retouched, and no AI disclosure is written onto the film, because there is no generated picture in it to disclose.
Narration is the one exception, and it is worth knowing before you switch it on. If you have MuseRig read a script over the film, the voice is generated even though the pictures are not, and disclosing that wherever you publish is yours to do. Leave narration off and nothing in the finished file was generated at all.
If what you want is a picture that does not exist yet, that is the Product photo route and it is a different job.
The order is the timeline
Add pictures from your Image shelf in the order you want them seen. Nothing re-sorts them — not by date, not by file name, not by shape. It is the hang you arranged, and it plays that way, top to bottom.
Click a picture on the shelf before you add it and it opens large, so you are choosing the photograph rather than a thumbnail of it. Once a picture is in the running order you can drag it where you want it, or use the up and down arrows on its row; both do the same thing and the arrows stay because dragging twenty-four rows is not always the faster way.
Up to 40 pictures, and up to three minutes. Forty is deliberately more than a show usually hangs: fourteen works with a close look at half of them is already twenty-one pictures, and the ceiling should not be the thing that decides how closely somebody may look at a painting. The length is the real limit.
A photograph too small is refused, not enlarged
Every picture is given the same box on the wall. The same box, so a small work does not swell to fill the frame and a large one does not shrink to match the last, and the whole set reads as hung rather than pasted. Inside that box a photograph is only ever made smaller, never larger, and always at its own proportions.
So a file that would have to be enlarged to fill its place is refused by name, with the number you need:
The photograph of "Vivarium I" is 800 by 600 pixels, which is smaller than the space it fills. MuseRig will not enlarge a picture, because enlarging softens the work. Send one at least 1019 by 764.
That is not a setting you can turn off. Enlarging softens a photograph, and a soft picture of somebody's work is worse than no film. Go back to the camera roll with the number.
Nothing is cropped either, which changes what the shape means here. Choosing a shape only changes how much wall is around the picture: a landscape photograph in a vertical frame is not a worse crop, it is a smaller picture on more wall. A set of landscape work belongs in landscape or square rather than vertical.
There are two tiers and no Quick. Quality is 1080p, for social and the web. Premium is 4K, for a screen in a room. The cheapest way to break the promise that your work arrives intact is to deliver it small, so the entry tier the other routes sell is not offered on this one.
The caption
Captions are set the way a gallery label is set: artist, then the title in italic with its year, then the medium, then the size. They are separate fields rather than one line you type, because that order is exact and getting it wrong is the fastest way for a film to read as amateur to the people it was made for.
Only the title is required. A builder's before-and-after has a title and no medium, and refusing to make the film over a missing year would be absurd.
The caption sits on a fixed line rather than under each picture's own lower edge, so it does not jump up and down as portraits and landscapes alternate. It fades in a second after the picture, so your eye reaches the work first.
It is centred under the picture by default. That is a correction rather than a preference: the first films set the label at the left margin, which is how a printed page sets a caption, and on screen under a centred painting it read as belonging to the wall rather than to the work. Left is still there and is right for some work — a wide landscape nearly filling the frame takes a left label well, and some artists label that way.
Your name does not have to be on every slide, and by default it is not there. A film that is one artist's work and opens with their name does not need to repeat it fourteen times. A group show turns it on, or names an artist per picture.
A detail says "Detail"
Mark a picture as a detail and it is understood as a close look at the work above it rather than as a work of its own. Its caption reads Detail and stops there: the medium and the dimensions are dropped.
That is not tidiness. "30 x 24 inches" printed under a crop of one corner states something that is not true of what is on screen. The dimensions belong to the whole work, so they stay with the whole work.
Set the length, not the pace
You can say it either way, and which way you choose decides what happens when you add a picture.
How long each picture holds. Two to twelve seconds; six is what a painting wants. The film is then however long that makes it.
How long the whole film runs. Eight seconds to three minutes. MuseRig works the holds out from it.
The second is the one people actually ask for out loud — make it thirty seconds — and it is the one that behaves when you change your mind. Add a detail shot to a film with a set length and every hold shortens slightly; the film is still thirty seconds. Add one to a film with a set pace and the film grows past the feed it was made for. The dissolve gives way too: on a short hold it shortens rather than swallowing the picture whole.
If the holds come out under four seconds MuseRig says the pace is brisk. It says so and makes the film anyway. Twenty pictures in thirty seconds is a legitimate thing to make, and refusing it would be MuseRig deciding your edit for you.
It refuses only when the arithmetic will not close, and then it says what you can have instead:
30 pictures will not fit into 30 seconds. The shortest this film can be is 42.6 seconds; drop pictures or ask for a longer film.
The same sentence arrives the other way round when a handful of pictures cannot fill a long film without holding each one past twelve seconds.
Before the quote
Five stages, each marked reviewed by you: Pictures, Captions, Cards, Sound and words and Review. The cards are the opening and closing text cards MuseRig letters itself, so the show's name and your website are spelled the way you typed them.
Each card takes a small line above, a headline, a small line below, a ground of ink, paper or your brand's accent, and one of your own photographs under the type if you want one. The headline is sized to fit rather than by counting characters, so a gallery's name stays on one line instead of breaking across two. Both cards preview in the shape the film is actually being made in, so a card you approve for a 4:5 cut is the card that gets made.
There is also a name for the top corner, and it starts empty. A picture film is usually somebody else's work, so nothing is stamped on it unless you type something — your own name, the gallery's, or nothing at all, which is what most exhibition films want.
Sound and words carries three switches that do not depend on each other: one music track from your Audio Library, narration read over the film, and a subtitle line per picture. Leave all three off and the film plays silent, with the gallery captions and nothing else, which is a perfectly good way to show paintings.
The quote refuses until all five stages are marked, and a connected agent cannot mark any of them.
From a connected agent
An agent can arrange the running order, write the captions, choose the shape and the tier, set the wall colour and the two cards, and say how long the film should be. get_plan_outlook then reads the plan back as it will actually play: every slide in order, each picture's title and the seconds it is alone on screen, the hold each one actually gets, the total, and the refusal in words when a length cannot be met. That is how an agent answers how long the film is and what is in it without guessing at either.
No picture in this film is generated. The pictures are yours, the order is yours, and MuseRig only holds them up.
The Library Audio shelf keeps useful music and sound effects available across projects without turning every file into a fake campaign.
MuseRig has one shared allowance for both kinds of audio: 50 files, 500 MB total, and 20 MB per file. The visible gauge shows both item and byte use. A refusal says which of the three stopped the upload, so a Library that is out of space does not read as a Library that is out of room for files. These limits keep the shelf purposeful and protect the workspace from clutter or accidental storage growth.
Each item keeps the title, creator, original source, license link, terms-check date, attribution requirement and a practical fit note. MuseRig stores the file privately and signs a temporary preview only for the signed-in workspace. Previews load on demand rather than while the whole shelf opens. Show the shelf as Cards, which carry all of that, or as a Grid of compact tiles with a play button and a Details button that opens the full card; the choice is remembered in this browser.
Confirming that you own a file, or hold a license that lets you keep it, is yours alone. A connected agent can research options, shortlist them and put a track on the shelf for you, and it cannot purchase a license, accept terms, confirm rights or delete anything.
Saving an item to the Library is not approval to spend or use it in every project. In a Faceless Video, the signed-in human can choose one Music bed and up to three Sound FX, place each effect on a five-second scene boundary, and confirm that the current licenses permit the intended use. Connected agents may suggest candidates, but they cannot attest rights, select files, place effects, or make the final decision.
A track your agent added
Ask a connected agent for music and it can put the track on the shelf itself, with the title, creator, source and license written exactly as the source states them. What it cannot write is that the license permits your use. That sentence carries your name, so only you can make it.
The track therefore arrives marked Added by your agent · rights not confirmed. Everything the agent wrote is there to read, and one button confirms it. Until you do, the track stays out of your films: the pickers do not offer it, and a quote for a project that names it is refused and says which track is waiting.
Open the license link before you confirm. MuseRig has checked nothing about it, and neither has the agent. It copied what the source says, which is the only honest thing it can do.
Mark the tracks you reach for first
Each music track on the shelf has 1, 2 and 3 under it. Press one to make that track your first, second or third choice, and the shelf moves it to the top with a label saying so. There is one of each: choosing a new first choice takes it off the old one, and MuseRig tells you which track that was. Press the number a track already holds to clear it.
Sound effects are not ranked. They are placed at a scene rather than chosen to sit under a whole film.
Which track went under which film is kept on the film, not on the shelf. So use the fit note to describe the music itself — its mood, its tempo, how it sits under a voice — rather than where it came in a shortlist. MuseRig will not let an agent write a ranking or a rights claim into a fit note, because a sentence written once and never reopened cannot keep either one true.
A track does not have to start at the beginning
Music written for listening opens quietly and often does not arrive until about thirty seconds in. Dropped under a fifteen-second film from its first frame, it sounds like the music was late rather than like the film began early.
In a synthetic spokesperson project you can set where in the track the bed starts, up to ten minutes in. That is a per-project choice made in the Sound stage rather than a property of the shelf item, so the same track can open at 0:00 in one film and 0:32 in another. Faceless Video still starts its bed at the beginning of the track.
One thing worth doing when you file a track away: use the fit note to record that the useful part starts a minute in. It saves listening through it again next time.
Keep source and license links with every reusable item. “Free” never means MuseRig should guess what a license permits.
MuseRig begins with a clearly synthetic presenter, illustrated or photoreal. A real person's likeness and cloned voices remain locked for a separate consent, revocation and provider-deletion acceptance.
The six review stages
Talent — Choose a saved synthetic presenter, or upload one private reference image, name them, and pick a curated preset voice. The name is suggested from the file you uploaded and is worth getting right: it names a reusable character you will pick by that name in every project afterwards, not just this video. A saved presenter arrives with the voice it was saved with, but that is a starting point rather than a fixed property: choose a different one and only this video changes. Changing the voice reopens this stage, so nothing is quoted with a voice nobody reviewed. Marking the stage reviewed saves a reusable MuseRig character record; it does not create a provider avatar or spend credits. Every presenter MuseRig can film is generated. Some are drawn and read as drawn; others are made to look like a photograph of somebody who does not exist. Say which on the Characters page, as Illustrated or Photoreal, and it shows on the card and reaches a connected agent choosing talent. It changes nothing about the film, and it never redraws the portrait: it is a note about the picture, never a claim that a person was photographed, and it can be changed at any time. A Real person character is a different thing entirely, and the only one that records a named adult's consent. Your account starts with four presenters, chosen from the ones MuseRig offers, and they arrive ready to film: a portrait, a voice and a working avatar, so there is nothing to switch on and nothing to pay before you use them. The four are yours rather than everybody's — which four you get is worked out from your account, so two businesses in the same town rarely open on the same faces — and they always include at least one illustrated presenter and at least one photoreal one, so you can see both kinds on your own shelf before choosing. They are yours to keep, rename or remove. A saved character's voice and reference can both be set on the Characters page, and a presenter needs both before their avatar can be built: choose the voice while saving them, or on their card afterwards, and hear it before you decide. Choosing a voice costs nothing and can be changed until the avatar exists, after which the presenter keeps the voice it was built with. A saved character's reference can be added or replaced on the Characters page; replacing the likeness of a presenter that already has a Runway avatar drops that avatar, so it is prepared again on the Create step, and reopens the Talent and Consent reviews on every draft that uses them.
Script — Review the exact words the presenter will say, choose how long the presenter gets — up to 15, 20, 30, 45 or 60 seconds — and choose the shape of the finished film and how the presenter meets it. Each length has its own reviewed word range, shown live as you type, because MuseRig speaks the whole approved script rather than trimming a sentence to fit. This stage also carries up to three cutaways.
Sound — Choose Narration only, Voice + my music, or Intentional silence. Music comes from your private, rights-confirmed Library, is normalized and ducked beneath the presenter voice, and can start anywhere in the track you choose. Only the signed-in human can select a track and confirm its current project-use license; choosing one spends no credits.
End card — Review the closing two seconds as they will render: the brand name, tagline, website and closing line, the ground colour, and one of your own images if you want one. The card is edited on the picture of itself: drag a line or the logo where you want it, or select it and use the arrow keys; the panel beside the card sets its words, size and colour, adds a line of your own or takes one off. The film plays exactly that picture. Any change to the card reopens this stage; a plan quoted before this stage existed keeps its review.
Consent — Choose End card only, On-video label, or No disclosure. The first two write the AI-generated presenter disclosure onto the video; No disclosure writes none, and the duty to disclose wherever you publish becomes yours. The project record names the presenter as synthetic either way. The signed-in human confirms that the reference is synthetic media they own, does not depict an identifiable real person, and may be used for this private presenter.
Create — Review the bound talent, voice, length, shape, presenter fit, sound, end card and disclosure treatment before the exact quote. A new presenter also needs a Runway avatar here: Save provider setup records a local checkpoint, then Create the avatar checks Runway and, once you confirm, builds the persistent avatar this presenter speaks with. Until that exists the presenter cannot be generated at all. Only the signed-in human can approve credits and start creation. The same Create avatar button is on the Characters page, on any presenter that has a portrait and a voice, so a presenter you are getting ready can be finished without starting a film for them.
The shape of the film
A spokesperson video used to be 9:16 and nothing else. There are now four destinations, the same four Faceless Video offers:
Social vertical, 9:16 — Shorts, Reels, TikTok.
Landscape, 16:9 — YouTube, websites and email.
Square, 1:1 — LinkedIn, Google and other feeds.
Feed portrait, 4:5 — Instagram and Facebook.
Choose this before you spend time on the end card and the cutaways, because it decides what every image in the film does. A 16:9 photograph framed inside a 9:16 film takes up about a third of the height, and cropping it to fill instead loses two thirds of the width. Landscape photographs want a landscape film.
Changing the shape is reversible draft work and spends nothing, but it reopens Script review and every stage after it, the End card included.
Fill or Fit
Runway generates the presenter, and MuseRig does not tell Runway what shape to send back — so the shape the presenter arrives in is not a number the app knows. That is the honest reason this control exists.
Fill crops the presenter to the frame. It is what every vertical film has always done and is proven there.
Fit shows the presenter whole, on your brand paper, and cannot crop them.
Fill is the default for Social vertical. Everywhere else the default is Fit, because cropping a shape nobody has measured into a frame it may not match is how a presenter loses the top of their head after the credits are spent. Changing the fit also reopens Script review.
How long the film actually is
The film is as long as the speech takes, not as long as the window you chose. The 15, 20, 30, 45 or 60 seconds is a ceiling. MuseRig speaks the whole approved script, and the picture ends when the presenter stops talking, plus the two-second end card.
The gap between the two can be large. A 55-word script approved in a 30-second window was delivered in 17 seconds. Pace is not a property of the voice either: the same preset has run at 128 words a minute in one script and about 200 in another. Punctuation is what slows text to speech down — every full stop, comma and dash adds a pause — so two scripts of the same length can differ by several seconds.
So MuseRig will not give you one number. The Script stage shows the fastest and the slowest plausible delivery for the words you actually wrote, and anything you time against the end of the film is checked against both.
Cutaways
A cutaway holds one of your own images while the approved voice keeps running underneath. You can place up to three in one film. Each one has a start — the sentence of the script it begins on, or a number of seconds before the end of the delivery — and a hold of one to eight seconds. However many you add, at least five seconds of the presenter stays on screen.
Framed shows the whole image with bands around it. Nothing is cropped, so a flyer keeps the edge of its own words.
Full frame fills the screen and crops to the shape of the film.
Inset keeps the presenter on screen and puts the image beside the presenter, already scaled without cropping.
An inset sits over the presenter, so it does not count against the five seconds the presenter is owed. It still takes exclusive possession of its stretch of the timeline: two cutaways never overlap, and one squeezed under half a second by the presenter floor or by its neighbour is dropped rather than rendered as a flash.
Placing a moment on a sentence
Say while the presenter says sentence 3, or the last sentence, and MuseRig finds those words on the delivery actually made. It works from the share of the words: in a 55-word script whose last sentence begins at word 51, the moment starts about 91% of the way through the speech, whether that speech takes 17 seconds or 26. Text to speech pauses at commas and full stops, so the start is an estimate of where the presenter is, not a transcript. Give a moment that must land on particular words a second or two of extra hold.
The workbench lists the script's sentences by number and shows the one each moment starts on. Edit the script and the moment stays on its sentence; if the sentence it points at is no longer there, the workbench says so and the moment is not rendered until you pick one that is. Seconds before the end is still there as the second choice, and every plan saved that way works exactly as it did.
What is checked before you pay
The delivery is the one thing nobody knows until the presenter has spoken. A moment placed in seconds is counted back from its end; a moment placed on a sentence moves with it, but still needs room beside the presenter floor and its neighbours. Each moment is therefore resolved twice — once against the fastest plausible delivery, once against the slowest — and the workbench tells you which of three it is:
It renders. It survives both.
It renders sometimes. It survives one and not the other, and nobody can tell in advance which you will get. Place it on a sentence, move it closer to the end, or shorten the hold.
It never renders. It falls outside the film however fast the presenter speaks, or names a sentence the script no longer has.
Here is why that matters. A moment placed 20 seconds before the end of a film that turns out to run 17 seconds starts three seconds before the film begins, so it is dropped and you have paid for an image nobody sees. Placed on the last sentence instead, the same moment starts when the presenter reaches those words, wherever they fall.
What is in the next chapter
The end card, where the music starts, copying a plan you have already reviewed, and what a connected agent may change are all in End cards, music and variants.
The closing card, where the music starts, and how to reuse a plan you have already reviewed. The six review stages, the shape of the film and the cutaways are in Review a synthetic spokesperson.
The end card
The closing two seconds are yours, and since v0.42.35 they are a review stage of their own, between Sound and Consent, so the card is approved before the quote rather than first seen in the film. The same card closes a product ad and a faceless video. Set the brand name, the tagline and the website, or leave them as they are for the MuseRig card. Any line you clear is not rendered at all. The stage draws the card at the shape of the film, and draws it as you work rather than a moment after you stop.
Since v0.42.36 the card is edited on that picture. Click a line or the logo and drag it where you want it; it snaps to the centre line and to the rows the classic card uses when it lands near one. The arrow keys move the selected element by a hundredth of the card, Shift by five hundredths, so nothing needs a mouse. The panel beside the card changes the selected line's words, its size (extra small to large), its colour (the ground's own ink, your Brand kit's colours, or any colour) and takes the line off; for the logo it offers a size stepper, a replacement from your shelf and removal. Add a line puts a line of your own on the card, up to three of eighty characters, for an address, a price or an opening time. Reset to the classic layout puts everything back. Lines keep clear of each other: a line moved onto another comes to rest just past it, in the direction it was going. The card you start from keeps itself clear too, so a long brand name that runs to two lines moves what is under it down rather than sitting on it. A layout that arrives with two lines through each other — from a connected agent, say — is not moved for you; the two are named under the card, and the stage cannot be marked reviewed until they are apart. Double-click a line, or press Enter with it selected, and you type into it where it sits, in the card's own type; Escape puts the words back. See it larger opens the same editor at the height of the window, with the handles, the keys and the panel all working there.
The card on the stage is drawn by your browser, which is why it moves the moment you do. MuseRig draws the card the film actually plays, from the same measurements, and See the film's own render under the card shows you that picture whenever you want it. The two are the same card: every position, size and colour comes from one place, and it is that, not either picture, which the review reads. So a line moved on the stage is moved on the film to the pixel.
The card's ground can be a colour of your own: the Review step offers the hex colours from the project's Brand kit as swatches first, then any colour, and Default returns to MuseRig's cream. The type goes pale on a dark ground, so it reads whatever you pick.
Switch the card to Image only and no type is rendered at all. Your wording is kept and comes back if you switch text on again. With no type there is no wash and no band reserved for it, so the image gets the whole frame.
You can also add one of your own images, and choose how it meets the frame:
Framed never crops. With text the image sits in the upper area on clean paper with the type below it; with no text it takes the whole frame. This is the right choice for a flyer, a poster, a logo — anything that already contains words.
Full bleed fills the card and crops to the shape of the film. With text it takes a paper wash so the type stays readable; with no text it is left alone. This suits a photograph with a calm area. It does not suit a document: the wash lowers contrast without removing detail, so the image's own words show through behind yours.
Most images are wider than they are tall, so in a vertical film full bleed usually crops more than people expect — often more than half the width. Framed keeps every edge.
The closing line is the quietest type on the card, which is right for an attribution and wrong for a phone number. So the card asks what it is holding. Leave the Made with love and creativity by MuseRig credit in place and it stays an attribution, small and grey. Put your own content there — an email address, a number, a booking link — and the line renders at contact weight instead, big enough to read off a phone.
Changing or removing the MuseRig credit is available once you have bought credits. Free-allowance credits are a grant, not a purchase. Until then the credit line can still be moved and resized like any other; only its words and its presence are held.
The end card and cutaways both draw on your Image shelf, so an image you added in one project is available in every project. The presenter's own reference is deliberately not offered: it is the likeness, not card art.
A cutaway can also be a text card instead of an image. You type a headline and, if you want them, a small line above and a small line below, and MuseRig draws the card itself in your Brand kit's colours: Ink is dark with pale type, Paper is the reverse, and Accent uses the kit's strongest colour. Nothing is generated and no credits are spent on it, so the words on screen are exactly the words you typed, which is what you want for a price, a date, an address or an offer. You can put one of your shelf photographs under the type, and the Review step shows the card as it will render. Shorter headlines are set larger, and one line up to ninety characters is the limit.
Where the music starts
Music written for listening opens quietly and often does not arrive until about thirty seconds in. Dropped under a fifteen-second film from its first frame, it sounds like the music was late rather than like the film began early.
So you can tell MuseRig where in the track to start the bed, up to ten minutes in. Nothing else changes: the bed is still normalized, still ducks beneath the presenter voice, and still resolves quietly on the end card. The start point only decides which part of the track you hear.
Making a variant
Make a copy on any project in your Library opens a new draft with the same settings — script, talent, voice, length, shape, presenter fit, sound mode, disclosure, end card and cutaways. Nothing is duplicated but the record itself: characters and images already belong to your workspace, so a copy costs no storage. Outputs, jobs, quotes and credits are never carried; a copy is a plan, not a result.
Two things are deliberately left behind, because they are statements you made about the original rather than settings: the synthetic rights confirmation, and the music track whose licence you confirmed for that project. A copy asks for both again, so it arrives needing the Consent stage and a track. Everything else is ready.
What a connected agent may do
A connected MCP agent may draft or revise the synthetic script, curated voice, presenter length, film shape, presenter fit, sound mode and direction, music start point, disclosure treatment, end-card wording, website and ground colour, and may name an existing image from your shelf for the end card or a cutaway, placed on a sentence of the script or in seconds before the end.
It may also choose the talent, from the presenters already saved in your workspace and ready to generate. That is the same permission that lets it name one of your images: it picks from a shelf you built. The presenter arrives with the reference and the voice you saved them with, exactly as when you pick them yourself, so a voice chosen earlier in the draft is replaced by theirs. All four review stages reopen, Consent included. The rights confirmation you already gave is not removed, and it is not moved onto the new presenter either: you confirmed one particular likeness, and nothing can carry that to another. You are asked again, against the presenter who is actually in the film, before anything is quoted.
It can save a presenter under a name, say whether they are illustrated or photoreal, and give them one of the eight curated voices, because neither says anything about a real person. It cannot attach their portrait: a character reference is a statement about whose face that is, so you add it yourself on the Characters page, and until you do the presenter cannot be quoted. It cannot record a real person's likeness, create the provider avatar, select audio, confirm rights, mark a human review complete, approve credits, or start generation. It can add an image to your shelf, which then waits for you to open it before any quote may name it. It can ask you to delete a saved presenter, and that request sits in your Library until you approve or decline it.
Before anyone pays, an agent can also read what the plan will do: the fastest and slowest plausible delivery for the words in the script, for each cutaway whether it renders, renders sometimes, or never renders, and the lines the end card will carry, in order, on which ground. Those are the three answers in Review a synthetic spokesperson, from the same check the workbench runs. The same answer lists the script's sentences by number and quotes the one each moment starts on, so an agent can place a moment on a sentence the way you would.
Agent revisions are merged field by field, so anything the agent did not send keeps the value you saved. Any field it does change reopens the review stages that depended on it — changing the script reopens sound and consent too, because both were reviewed against those words, and changing the shape or the fit reopens everything from Script onwards.
Narration-only remains the safe default. A music-enabled render is a separate paid creation and still requires the exact quote and explicit human approval.
