Read the settings as a system
The graph uses a dedicated MiniMaxH3TurboLoRA node and separate video and audio sigma shifts of 12 and 3. The adapter, loader, scheduler and encoders all contribute to the result. Removing the adapter while leaving a very low step count is not equivalent to the configured turbo workflow.
The current local environment selects a pruned W4A8 FL2VA model and a 4B text encoder with a matching projection. Older downloadable graph examples may show earlier loaders. Treat those examples as versioned snapshots and check file compatibility before reproducing them locally.
Reproduce carefully
- Start with the existing downloadable JollyAI H3 workflow guide and note its snapshot date and dependencies.
- Verify the model and encoder names against the graph you intend to run.
- Change one setting at a time and keep the original working graph.
- Judge timing and output quality together on the same small test set.
What six steps does not prove
It does not establish that six is best for every checkpoint, that more steps can never help or that all six-step services have equal quality. Distillation and workflow changes alter the tradeoff. The public playground intentionally abstracts these engineering settings so users can focus on the shot.
For creators using the hosted tool
Start by improving the source image, motion brief and composition. If an output fails repeatedly, record the workflow and visible problem for support. Increasing a number mentioned in a prompt does not change server-side sampling steps. Use the current interface for available controls and the technical guide for implementation context.
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.