Measure three intervals
Queue time runs from submission to the start of processing. Processing time includes the work the service performs after that point, which may include loading, encoding and output handling. End-to-end time runs until the completed file is available. None should be labelled pure sampling time unless you have measured that specific stage.
A single successful sample is an observation, not a fleet-wide benchmark. Warm model state, other work and the selected output settings can change the result. A duration shown by a browser player is the length of the video, not how long generation took.
Keep a reproducible record
- Record the workflow, prompt, input files, requested duration and resolution setting.
- Record submission, processing start and completion timestamps when available.
- Inspect the final file's dimensions, frame rate, duration and audio streams.
- Repeat a defined test set before calculating a median or making a comparison.
Compare quality as well as time
Use the same acceptance checklist for every result: motion continuity, subject fidelity and usable sound. Report failed attempts too. Timing only the best successful clip hides retry costs and can make a workflow look more efficient than it is for real production.
How to read our examples
The guide demonstration is a selected owner-generated sample with its request and file metadata recorded. It is not a matched comparison against another provider. Where processing timestamps are unavailable, we show no speed claim. For a deadline-sensitive project, leave room for queue variation, review and editing rather than scheduling around one observed render.
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 ↗ORIGINAL JOLLYAI GENERATION
A five-second request, inspected as a real file
- Request
- 5 seconds / 0.4 MP
- Delivered file
- 864 × 480 / 24 fps / 5.167 s
- Audio
- Generated audio stream
- Queue wait
- 22 seconds
- Processing interval
- 115 seconds
- End to end
- 137 seconds
One owner-account observation, September 20, 2026. Processing is the recorded job interval, including pipeline overhead, not isolated sampling time. Other requests and settings can differ. Five seconds was an administrator request; public H3 presets currently use 4/6/8/10 seconds.
Exact demonstration prompt
One continuous studio product film of a fictional unbranded acid-green sports shoe resting on a matte charcoal pedestal. The camera slowly slides from left to right through a modest angle while the shoe remains stationary. A thin veil of pale mist drifts behind the pedestal. A narrow white light sweeps across the fabric texture once, revealing the mesh and sole. Clean black background, sharp silhouette, believable proportions, premium commercial photography, no writing or logos. Sound: a soft airy whoosh and quiet studio ambience. Landscape 16:9.
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.