MovieFlow Video Agent: from one story prompt to a finished film
MovieFlow Video Agent is the agent mode in the MovieFlow home composer that takes a story you describe in plain language and produces a complete short film: it outlines the story, shapes the characters and scenes, arranges the shots and their rhythm, generates every shot, adds background audio, and assembles a final cut you can open and edit. You choose Video Agent from the mode switch above the composer, type the movie you want, and the agent does the production work that a small crew would otherwise split between a writer, a storyboard artist, and an editor.
The name matters because MovieFlow has more than one way to create. AI Video makes a single clip. Agent V2 is the conversational assistant. Video Agent is the mode built for a whole film from one idea, and it is the mode most people mean when they search for MovieFlow Video Agent or the MovieFlow AI agent for movies. This page explains what it does, what it needs from you, and where the finished project goes.
How to make a film with Video Agent
Video Agent lives on the home page. You do not install anything or open a separate editor first. The steps below follow the order the product shows on screen.
- Open the MovieFlow home page and sign in. Video Agent needs a Plus or higher membership, so check your plan first.
- Switch the composer mode to Video Agent. The placeholder changes to "Describe the movie story you want to create".
- Write the story: who it is about, where it happens, what changes by the end. Turn on Enhance script if you want the agent to expand a short idea.
- Add up to three reference assets if you have them. Video Agent accepts images and audio, such as a character still or a voice line.
- Send the prompt and let the agent work. Progress messages show each stage, from outlining the story to generating the final product.
- Open the finished project in the workflow editor, replace any shot that misses, then export the film you want to keep.
What the agent does between your prompt and the cut
The agent does not jump straight from text to one long video. It works in stages, and the screen reports each one. First it outlines the story and writes the script and prompts. Then it shapes the characters and the scene details, draws a sketch for each character, and drafts a sketch for every shot. After that it arranges the shots and the rhythm, polishes the dialogue and transitions, and generates the video for each shot. Background audio is generated next, and a post-production pass stitches everything into the final product.
That order is the reason a Video Agent film holds together better than a pile of separate clips. Characters are drawn before shots are rendered, so the same face can carry from scene to scene. Dialogue is polished before the edit, so lines land where the story needs them. When a stage fails, the agent retries the task before it reports a problem, and you see how many retries it used.
A typical run tells you how long to wait. Shorter stories report that the film should be ready in about three to four minutes, and longer ones in about six to seven minutes. You do not need to keep the tab in focus. The project is saved, and you can do other work while it renders.
Story prompts that work well
Example: "A retired lighthouse keeper teaches his granddaughter to read the storm signals, and on the last night the lamp fails and she has to light it alone." This prompt names two characters, one place, and a turn at the end. The agent can split it into an opening, a lesson, the failure, and the resolution, and each of those becomes a group of shots.
Example: "A delivery robot gets lost in a night market and is guided home by a street cat." Here the characters are not human, which is a good test of character consistency. If the robot changes shape between scenes, replace those shots in the editor instead of starting over.
Example with a reference: upload a portrait as Image 1 and write "Image 1 is a violinist who plays one last concert in an empty hall before it is torn down." The still gives the agent a face to keep. Audio works the same way: attach a short voice line and describe where it should be heard.
These are prompts you can run yourself, not published MovieFlow case studies, and they do not come with view counts. Finished films from other people are in AIdeo World, the community feed, if you want to see pacing before you write.
Editing the result in the workflow editor
Every Video Agent film opens as a project with two views: Script and Storyboard. The Script view keeps the story text and the narration for each scene. The Storyboard view shows the shots in order, each with its shot script, its first frame and last frame, and its dialogue.
The timeline below the preview is a working editor, not a static player. You can split a clip at the playhead, delete a segment, add a new shot, extend a shot, undo and redo, choose a voice for narration, load subtitles onto the timeline, show or hide them in the preview, add background music, merge all storyboards, download a single shot, and download the videos. Keyboard shortcuts follow common editors: Space plays and pauses, Ctrl+S splits at the playhead, and Ctrl+Z undoes.
When a shot needs a different result, you can edit the frame, edit the clip, add a shot from a text prompt, from an uploaded reference frame, or from assets in your library, or replace the character in a video with another person while keeping everything else unchanged. Applying a structural change regenerates the storyboard and the steps after it, and the editor asks you to confirm before it does that.
Models behind Video Agent
MovieFlow is a multi-model platform. The home page lists the video and image models that are live, including Seedance, Kling, Vidu, Veo, MiniMax H3, and MovieFlow's own models, plus image models such as Nano Banana and GPT Image. Video Agent matches a model to each shot type, so a dialogue close-up and a wide establishing shot do not have to come from the same generator. The exact lineup changes as models are added, so this page does not freeze a list.
The platform core, agent orchestration, and production pipeline are built by MovieFlow. Generation connects to leading models through MovieFlow's own matching layer. That is the practical difference from a single-model tool: you describe the film once, and the agent decides which model fits each shot, while you can still step in and change any step.
Plans, credits, and what is charged
Video Agent requires a Plus or higher membership. If your credits are not enough for a creation, the product tells you before it starts and offers the upgrade path. No credits are deducted until a video is generated successfully, so a failed render does not cost you a finished film. Live plan prices and credit amounts are on the pricing page, and this page does not print them because they change.
Videos made on paid plans include commercial rights for ads, social media, and marketing, subject to the Terms of Use and your subscription agreement.
Video Agent questions
What is MovieFlow Video Agent?
It is the agent mode in the MovieFlow home composer that turns one story prompt into a scripted, storyboarded, and edited short film. It writes the script, draws the characters, generates each shot, adds background audio, and assembles the cut.
Is Video Agent free?
Video Agent needs a Plus or higher membership. New accounts receive free credits to try the core workflow, and other modes such as AI Video can be used to learn the tool first. The pricing page lists the current plans.
What can I upload to Video Agent?
Images and audio, up to three assets in the composer. Use an image for a character face or a location, and audio for a voice line or a sound you want in the film.
How long does a Video Agent film take?
The progress screen gives an estimate for each run. Shorter stories report about three to four minutes and longer stories about six to seven minutes. You can leave the page while it renders.
Can I change a shot after the film is made?
Yes. Open the project in the workflow editor, then edit the frame, edit the clip, extend the shot, add a new shot, or replace the character. The rest of the film stays as it is.
How is Video Agent different from Agent V2?
Agent V2 is a conversational assistant that plans and runs generation jobs step by step inside MovieFlow Studio. Video Agent is the home composer mode that produces a whole film from a single story prompt.
Where Video Agent fits in MovieFlow
Use Video Agent when you have a story and want a film back. Use AI Video when you only need one clip with direct prompt control, first and last frames, or omni references. Use MovieFlow Canvas when you want to wire images, videos, and audio together by hand. Use Feature Film when the story is longer and you want to approve one budget for the whole production. MovieFlow Studio holds the projects, assets, and team settings behind all of them.