Match movement to the purpose
Use a slow push-in when the viewer should notice one detail. Use lateral tracking when the subject travels through a scene. Use a restrained orbit when seeing another side of an object matters. An orbit asks the model to invent hidden surfaces, so it is a more demanding starting point than a small push-in.
Describe the subject's motion separately. “The camera tracks the bicycle” is incomplete if the bicycle's direction and pace are unspecified. A stationary object and a moving camera are often easier to judge than a scene where everything accelerates.
Run a useful camera test
- Write a fixed scene with one subject and a plain background.
- Generate a locked-camera version to establish whether the subject itself is stable.
- Change only the camera sentence for the next attempt.
- Compare background displacement, framing and object shape, not just how dramatic the clip feels.
Three starting instructions
A ceramic cup remains still on a wooden table. The camera makes a gentle straight push-in toward the handle. The horizon stays level and the cup remains fully visible. Soft window light, one uninterrupted shot.
For a pan, replace the movement sentence with “The camera turns slowly right from a fixed position, revealing the rest of the table.” For tracking, use a genuinely moving subject and state whether the camera stays beside it or follows behind it.
When the result feels chaotic
Remove simultaneous zoom, orbit and handheld instructions. Ask for a smaller movement and provide more space around the subject. If a specific starting composition matters, switch to image-to-video. Camera vocabulary guides a generative model; it does not provide the precise path controls of a 3D animation rig.
REVIEWED JOLLYAI IMPLEMENTATION
The workflow behind the guide
Written scene → text conditioning → turbo audio/video sampling → separate video and audio decoding → combined video file.
- Inputs, presets & access
- Limited free access; four-second free preset, with six/eight/ten-second PRO options. Current ordinary pixel budget: 0.4 MP. Use the account controls for current availability.
- Implementation note
- Local H3 turbo workflow. Six sampling steps; simple scheduler; turbo adapter; video/audio shifts 12/3. Current environment: pruned W4A8 FL2VA weights with a 4B encoder and matching projection. R2V uses a separate reference checkpoint. These settings describe the reviewed JollyAI deployment, not every H3 implementation.
Configuration reviewed September 20, 2026. Your current playground controls and plans page determine availability. Prompt recipes above are suggestions, not guarantees or claimed test results.
Open Text to video ↗Sources and useful next steps
The implementation notes describe JollyAI's reviewed local integration. Upstream capabilities and licences may differ from the settings available here.
- MiniMax H3 upstream project and model information ↗
- MiniMax H3 tool overview on JollyAI
- Earlier H3 ComfyUI workflow snapshot and downloadable JSON — check the newer configuration notes above when reproducing it.
- JollyAI content policy and service terms
Found an outdated setting or a reproducible problem? Contact JollyAI with the workflow and job reference. Keep account tokens and private input files out of public reports.