← Meta Muse for Filmmakers

Free lesson · lesson 1 of 23

What Muse Is and Isn't

Three things called Muse

Before you set anything up, be sure you're looking at the right product. Searches for "Muse" turn up several different things.

  • Muse is Meta's personal AI agent, at muse.ai. It launched in the US on September 8, 2026 and was built by Meta Superintelligence Labs. This course is about Muse.
  • The Meta AI app is a separate Meta product with its own terms. Settings you change in one don't apply to the other.
  • Muse Spark is the model, meaning the underlying AI engine, that Muse runs on. You don't sign up for Muse Spark. You use Muse, and Muse uses Spark.
  • Unrelated products share the name, such as Project MUSE (academic journals) and a website builder called Muse.

Here's why it matters on a production. Say your post supervisor forwards a "Muse setup guide" that turns out to describe the Meta AI app. You could switch off training in the wrong app and believe your notes are protected when they aren't. In this course, when a menu name or setting comes up, it refers to Muse at muse.ai and nothing else.

Try itMatch each name to what it is

Tap one on the left, then what it goes with on the right.

0 of 4 matched

Only one of these is the agent this course teaches.

Try it

A colleague says, "I already turned off training in Meta AI, so Muse is covered." What's the right response?

An agent acts; a chatbot answers

An AI agent is software that carries out tasks for you rather than only answering questions. Muse can search the web, fill in forms, draft and send email, and, with your approval, book or buy things. You'll use only a narrow part of that. In this course Muse won't send, sign, book or pay for anything.

Three features make it different from a chat window:

  • Background goals. You give Muse a long-term aim, it builds a plan and keeps working after you close the app. It comes back when something changes or when it needs your approval.
  • Memory. It remembers what you tell it (your role, your production, your preferences) and can make suggestions without being asked. You can tell it to forget things.
  • Approvals. Before certain actions it shows an approval card that you accept or deny.

For example, a chatbot will tell you what a festival tracker might contain. An agent can build the tracker, check the official festival pages every week and remind you before each deadline. That's useful, and it's also why the rest of this module is about limits.

See itHow a Muse task runs

Your approval card is the point where you decide what happens next.

Try it

Which of these is something Muse does that a question-and-answer chatbot doesn't?

Its own computer, Artifacts and connectors

Muse doesn't run only on your phone. Each user gets a Muse Secure VM, a cloud computer of its own with a web browser, a file system and a terminal. Muse can write code and build small tools there for a task.

What it makes is often an Artifact: a document, PDF, web page or interactive output such as a tracker or checklist. A festival tracker or a draft call-sheet layout would be an Artifact.

To reach your other services it uses Connectors, which are links to apps such as Google Calendar, Gmail, Slack, Box and Dropbox. Once connected, Muse can read from them and, depending on your settings, act in them. On a Mac, the Muse app can also use local apps and files, which Meta calls computer use.

Each of these is a way in to your production's material, so each needs care. Later lessons set connectors to read-only where that's offered, limit Mac access to one project folder and keep Artifacts unshared. For now, just know that these parts exist.

Try itMuse's working parts, in production terms

Flip each card to see what it means for a production office.

Try it

Muse produces a festival deadline tracker you can open and reuse. What is that called?

Agents carry the burden, not the art

This course gives Muse a clear job description. Ben Affleck, the Oscar-winning filmmaker who founded the AI tools company InterPositive, has said AI will fill in the parts of production that are "expensive and burdensome", and that it will always rest on human artistry. Here, that means the burden goes to the agent and the art stays with you.

Paperwork an agent can carry (as a first pass that a person then checks):

  • A first-pass shooting schedule built from a scene list with no dialogue
  • A list of scenes touched by new revision pages
  • Permit research from official film-office pages
  • Festival deadlines and reminders
  • A notes list for one named cut

Calls that stay with people:

  • Casting, performance and creative choices
  • Which draft is the current script (the script supervisor confirms)
  • Safety, availability and cost commitments
  • Anything involving a performer's face, voice or consent

For example, Muse can point out that scene 42 needs the house keys on the porch rail and scene 44 has to match. Whether to add an insert or keep the take is a decision for the director and department heads.

See itPaperwork versus calls

Muse can carry (first pass)

  • First-pass schedule from a dialogue-free scene list
  • Scenes touched by new revision pages
  • Permit research from official sources
  • Festival deadline tracker with reminders
  • Notes list for one named cut

Stays with people

  • Casting, performance and creative choices
  • Which draft is the baseline
  • Safety, availability and cost commitments
  • Any use of a performer's face or voice
  • Signing, booking or paying

Muse drafts the left column. People decide the right column.

Try it

Which task belongs with a person rather than with Muse?

Every task has edges and stop points

Richard Chuang co-founded Pacific Data Images, which became part of DreamWorks Animation, and built the pipeline behind Shrek. He has described a studio asking him for a fully automatic pipeline that would update the script, the breakdown and the generated material on its own. He called his own sketch of it rough and warned against taking such a pipeline for granted.

This course takes that warning seriously. Every standing task you give Muse has edges:

  1. What wakes it: a schedule, or an event such as new pages arriving.
  2. What it may change: usually only a draft or a list of its own.
  3. What it must never touch: the script, the cut, crew records, anything you send or pay for.
  4. Whom it asks: the person who makes the call.

A stop point is a named moment when Muse has to pause and wait for a person. In the revision check you'll build later, if two drafts arrive and one has a newer date but older content, Muse stops and asks the script supervisor which is current. It doesn't guess. Stopping at the right moment counts as the task working.

Try itMatch each edge to the revision-check example

Tap one on the left, then what it goes with on the right.

0 of 4 matched

These four edges make up the outline of every standing task in this course.

Try it

In the weekly revision check, a duplicate of scene 42 arrives alongside a draft with a confusing date. What should Muse do?

Checkpoint: confirm you're on the real Muse

Goal: make sure muse.ai is Meta's Muse before you enter anything.

You need: a browser on a computer or phone. Don't create an account yet. The next lessons cover a separate work account and turning training off before you start.

Steps:

  1. Type muse.ai into the address bar yourself. Don't use a link from an email or a search ad.
  2. Watch the address bar. Sign-in should pass through auth.muse.ai to a Meta login (you may see facebook.com).
  3. Stop at the login screen and close the tab.

It worked if the login page is Meta's. The screen may look different from any description here, because Meta changes the app often. What matters is that the login is Meta's.

If it doesn't work:

  • A login page that isn't Meta's isn't Muse. Close it.
  • Unofficial WhatsApp numbers or links that claim to be Muse should be treated as scams.
  • If Muse says it isn't available, check where it's offered (the next lesson covers this). Don't use a VPN or a foreign app-store account to get around the limit.
Try itPut the checkpoint steps in order

Put these in order: tap two to swap them, or use the arrows.

  1. 1
  2. 2
  3. 3
  4. 4

Stop at the login screen. Setting up the account comes later.

Try it

You open a link called "Muse for Filmmakers" and get a sign-in page that isn't a Meta login. What do you do?

In short

Muse is Meta's personal AI agent at muse.ai. It launched on September 8, 2026, was built by Meta Superintelligence Labs and runs on the Muse Spark model. It is separate from the Meta AI app and from unrelated products called Muse. Muse carries out tasks instead of only answering: it works on goals in the background, keeps a memory, builds Artifacts on its own cloud computer (the Muse Secure VM) and reaches other services through Connectors, plus local apps and files on a Mac. This course follows Ben Affleck's point that agents carry the burden, not the art. Schedules, revision lists, permit research, festival dates and notes lists are paperwork for Muse. Creative, casting, safety and baseline-script calls stay with people. Following Richard Chuang's warning not to take automated pipelines for granted, every task has edges and named stop points. The real Muse always signs in through a Meta login.

Carry on with the whole course

22 more lessons, a tutor you can ask anything in every one, quizzes, flashcards and a certificate. $97.

See everything it covers

Not ready yet? Have this lesson emailed to you.

Email me this free lesson and occasional news about this course. Unsubscribe any time with one click.