MovieFlow Canvas: a visual workspace for AI images, video, and sound

MovieFlow Canvas is a node-based creative workspace inside MovieFlow Studio. Each node is one piece of media, such as an uploaded file, a generated image, a generated video, or generated audio, and you connect an output port of one node to an input port of another to build a creative workflow. A built-in assistant, Canvas Agent, reads what is on the canvas and can generate or edit images and videos, extract or insert frames, replace frames, and apply exact trims when you ask it in plain language.

Canvas is the right tool when a film is not a straight line from prompt to cut. A product launch might need four stills, two motion tests, a voice track, and a final edit that pulls from all of them. On a canvas those pieces sit side by side, the links between them are visible, and you can rerun one branch without touching the others. On the MovieFlow home page this mode is listed as Canvas: a visual workspace for freeform editing and video refinement.

How to start a project on MovieFlow Canvas

You can reach Canvas from the creation cards on the MovieFlow home page. Signing in hands your session to MovieFlow Studio, where the canvas opens.

  1. On the home page, choose the Canvas card under the creation options and sign in if asked.
  2. Start from an existing project, or pick Free Canvas to begin on a blank canvas that is independent from your other projects.
  3. Open the node palette and add the first node: Upload to bring in a local file from the asset library, or Image, Video, or Audio to generate from a prompt.
  4. Connect an output port to an input port. A connected node uses the upstream result, for example a generated still as the start of a video.
  5. Type a request to Canvas Agent. Press @ to mention an asset and / for commands, then review the plan before anything is generated.
  6. Keep the results you like in the project library, then continue the edit later or move the material into a film.

Nodes and connections

The node palette has four starting points. Upload opens the asset library so you can add a local file. Image generates an image from a prompt. Video generates a video from a prompt. Audio generates voice, music, or sound effects. From any node you can create a downstream node, so a still can continue into a new image or into a video without copying files around.

Nodes keep their own tools. After an image is generated you can erase part of it. After a video is generated you can extract its first frame and send that frame onward. After audio is generated you can download it. Prompts sit inside the node, and you can drag a corner to resize the prompt when it gets long. Double-clicking a node puts it into the Agent conversation, which is the fastest way to say "change this one".

Because every link is visible, Canvas works as documentation too. A teammate who opens the project can see which still fed which video, and which version of a shot came from which prompt.

What Canvas Agent can do

Canvas Agent is a chat panel attached to the canvas. You describe what you want, and it turns the request into work on the nodes. It can generate new images and videos, edit images, edit videos with up to four reference images, extract frames from a video at a timestamp you choose, insert or replace frames, and trim a clip exactly. When a request needs many outputs, the agent drafts a batch plan with one row per task. You can ask it to revise all rows or only the selected ones before you approve.

There are two ways to run it. In the default mode, the agent waits for your confirmation in the conversation before every generation action, so nothing is spent without a click. In free mode, the agent plans generation tasks and executes them automatically. When a batch starts, video nodes run independently, so one slow shot does not hold up the rest, and results return to the canvas as they finish.

If an upstream model cannot complete a request, the agent says so and suggests trying again later, switching model, or writing a clearer description. If many reference images are attached, it asks you to give each image one clear role, such as subject, environment, or composition. Those messages are the product's own guidance, and following them is usually faster than resubmitting the same prompt.

Canvas layouts people build

Product spot: upload one packshot, add two Image nodes for lifestyle stills from that packshot, then add a Video node from each still. Ask Canvas Agent to trim both videos to the same length and extract a clean first frame for the thumbnail.

Character test: generate a portrait, connect it to three Video nodes with different actions, and compare the motion side by side. Keep the one that holds the face and delete the others.

Music video sketch: add an Audio node for a short track, then branch three Video nodes for verse, chorus, and bridge. The canvas keeps the structure clear before you commit to a full edit.

These are layouts you can build yourself. They are not published MovieFlow case studies. Finished community films are in AIdeo World if you want references for pacing.

Models on the canvas

Each generation node lets you choose the model that suits it, from the same multi-model catalog MovieFlow lists on its home page, which includes Seedance, Kling, Vidu, Veo, MiniMax H3, and MovieFlow's own video models, plus Nano Banana, GPT Image, and Seedream for images. Different nodes in one canvas can use different models. The live list and the per-model rates are in the product and on the pricing page, so they are not copied here.

Free Canvas and project storage

Free Canvas starts on a blank canvas, and the assets stay in that new project. It is independent from existing projects, generated media goes to that project's library, and you can continue later in Director Mode. That makes Free Canvas a safe place for experiments: nothing you try there changes a film you are already working on.

Generation on the canvas uses credits from your account. With the default confirmation mode, you approve each generation in the conversation. Downloads of finished work need a paid plan or a top-up. The pricing page explains the current plans.

Canvas Agent requests you can copy

Requests work best when they name the node and the outcome. "Make three variations of @Image 1 with a warmer evening light" tells the agent which asset to start from and what should change. "Extract the frame at two seconds from the selected video and use it as the start of a new video where the camera pulls back" chains two actions in one sentence, and the agent turns it into a frame task followed by a video node.

For edits, say what to keep. "Trim the selected video to the moment the door opens, keep the sound" is clearer than "shorten this". For batches, give the count and the rule: "Create a video for each of the four product stills, same camera move, five seconds each". The agent drafts four rows you can check before anything runs, and you can ask it to revise only the second row if that one needs a different move.

When the agent cannot help with a request, it says what it can do instead: generate and edit images or videos, extract frames, and apply exact trims. Select the relevant media first, or describe the media and the outcome you want, and try again.

Canvas compared with a timeline

A timeline answers the question "what plays when". A canvas answers "where did this come from, and what else did it feed". That is why Canvas suits the stage before the final edit: look development, product variations, character tests, and any job where one source image turns into many outputs. When the material is ready, the order of the film still matters, and that is where the Workflow editor or an export to your own editor takes over.

MovieFlow Canvas questions

What is MovieFlow Canvas?

A node-based workspace in MovieFlow Studio where uploaded files, generated images, videos, and audio are nodes you connect into a workflow, with Canvas Agent available to generate, edit, extract frames, and trim on request.

Do I need to know node editors to use it?

No. You can type a request to Canvas Agent and let it create the nodes. The connections are there when you want more control.

Does Canvas Agent spend credits without asking?

Not in the default mode. The agent waits for your confirmation in the conversation before every generation action. Free mode is the option that executes automatically.

What is Free Canvas?

A blank canvas that creates a separate project. Generated media goes to that project's library, and you can continue later in Director Mode without affecting other projects.

Can I edit a video I already generated?

Yes. Select the video node and ask Canvas Agent to trim it, extract or replace frames, or edit it with up to four reference images.

When to use Canvas instead of an agent film

Choose Video Agent when you want a whole film from a story prompt. Choose Canvas when you want to see and control how each asset was made. Many projects use both: the agent produces a first cut, and the shots that need care move to a canvas where you can branch, compare, and trim them.