Generate videos from text in Claude and ChatGPT

Claude can't generate video by itself. Connect Makemancer and Claude, ChatGPT or Codex can turn a written description into a short clip with sound, using Gemini Omni Flash (the default), Wan 3.0, Seedance 2.0 Mini, MiniMax H3, LTX 2.3 Fast or HappyHorse 1.0. A 5-second 480p Seedance 2.0 Mini clip costs $0.15.

Bring
A sentence describing the shot
Get
An MP4 clip, with sound on most models
From
15 credits · $0.15

Made with it

Our own results, first takes. Each caption says what it took.

Fig. 1A paper boat drifting down a rain-soaked gutter at night, neon in the waterWan 3.0 · 480p · 16:9 · 33 credits
Fig. 2Hands shaping a clay bowl on a pottery wheel in warm window lightMiniMax H3 · 768p · 16:9 · 96 credits

How to do it

  1. Connect Makemancer

    Add Makemancer to Claude, ChatGPT, Claude Code or Codex. The steps for each client are on the connection page.

  2. Describe one shot

    Write the scene as a single shot: subject, action, camera, light and sound. Give the shape and length, and ask for a quote before it runs.

    Make a 16:9 video: a low close-up of a small white paper boat drifting down a rain-soaked city gutter at night, neon signs reflected in the rippling water, raindrops hitting the surface, shallow depth of field. Use Seedance 2.0 Mini at 480p and tell me the cost first.
  3. Change one thing, then render the final

    Keep the prompt that worked and change one detail at a time. When the draft is right, ask for the same shot at a higher resolution or on another model.

    Keep the same shot but make the rain heavier and slow the boat down. When I approve it, render it at 720p.

Which model to use

ModelBest forPriced forCredits
Gemini Omni 1.1 Flashvideo/gemini-omni-flashThe default: fast generation from text with native audio.720p · 4 s40$0.40
Wan 3.0video/wan-3.0Native audio, 2-30 s from 480p to 1080p; can take a document or web page as the brief.Opens after your first credit purchase480p · 5 s33$0.33
Seedance 2.0 Minivideo/seedance-2.0-miniLow-cost drafts, 4-15 s at 480p or 720p with sound.480p · 5 s15$0.15
MiniMax H3video/minimax-h3Higher-resolution output at 768p or 2K, 4-15 s, with native audio.768p · 4 s64$0.64
LTX 2.3 Fastvideo/ltx-2.3-fastLonger 6-20 s clips at 1080p with native audio.Opens after your first credit purchase1080p · 6 s72$0.72
HappyHorse 1.0video/happyhorse-1.0Audio and video made together with lip-sync, 3-15 s at 720p or 1080p.Opens after your first credit purchase720p · 5 s98$0.98

Write the prompt like a shot list

  • One shot per request. Subject, action, camera position and light. "Low close-up, the boat drifts, neon reflections" is a shot; a whole story is not.
  • Name the camera move. Static, slow push-in, pan left or handheld. Without one, models pick their own.
  • Describe the sound. Rain, a quiet studio, distant traffic. The audio follows what you write.
  • Keep it slow. Calm motion holds together better than fast action, especially with hands and faces.

Settings that matter

Length and resolution set the price. Draft at the lowest resolution your model offers (480p on Seedance 2.0 Mini, 768p on MiniMax H3), then render the take you keep at a higher one. Ask for the aspect ratio you need: 16:9 for the web, 9:16 for stories and reels. When you already have a still you like, image to video gives more control than text alone.

What we got

The paper boat clip came from Wan 3.0 at 480p, one of the models that opens after your first credit purchase; the same prompt runs on Seedance 2.0 Mini before that. The pottery wheel, hands shaping a clay bowl in warm window light, came from MiniMax H3 at 768p. Both prompts are in the examples linked on this page.

Questions

Can Claude generate videos?

Not on its own. With Makemancer connected, Claude writes the request, sends it to a video model and gives you the finished clip with a permanent link.

Is there a video generator in ChatGPT I can use with other models?

Yes, if your ChatGPT plan and workspace allow custom MCP servers. Add Makemancer from Plugins and ask for a clip with Wan, Seedance or another model by name.

Which model should I start with?

Start with Seedance 2.0 Mini at 480p or Gemini Omni Flash to test an idea cheaply, and use MiniMax H3 for a sharper final take. Wan 3.0, LTX 2.3 Fast and HappyHorse 1.0 open after your first credit purchase; the table marks them.

Do the videos have sound?

Yes. All six models in the table generate audio with the picture. Describe the sound you want, such as rain on water or a quiet room.

How much does a text-to-video clip cost?

From $0.15, depending on model, length and resolution. A 5-second 480p Seedance 2.0 Mini clip costs $0.15, and the assistant quotes each request before it runs.