VIDEO GENERATION API

H3 Max API

fal H3 Max text-to-video with stronger prompt adherence, polished aesthetics, and fast hosted inference. Route it through one ManifoldGen key and receive a recoverable asynchronous job.

Authentication

Bearer API key

Response

Async job + status URL

Request schema

servicevideo_generateRequired

modelminimax/h3-max/text-to-videoRequired

promptstringRequired

image_urlHTTPS URLOptional

duration5 | 10 | 155

resolution480p | 768p480p

aspect_ratio16:9 | 9:16 | 1:1 | 4:3 | 3:4 | 21:916:9

Job lifecycle: POST to /api/service, read result.job_id, then poll /api/video-jobs/{job_id} every 2–5 seconds. Completed jobs contain a durable video_url.

Open the visual generator

Live API tester

Generator controls

Describe the shot

DurationAspect ratioResolutionSeed

Output

Your generated shot will appear here, ready to download or place on the Studio timeline.

API request

curl https://manifoldgen.com/api/service \
  -X POST \
  -H "Authorization: Bearer $MANIFOLDGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "service": "video_generate",
  "model": "minimax/h3-max/text-to-video",
  "prompt": "A cinematic tracking shot through a rain-soaked night market, natural motion, rich reflections",
  "duration": 5,
  "aspect_ratio": "16:9",
  "resolution": "480p",
  "output_format": "mp4",
  "include_audio": false
}'

The API and this tester use the same request. Provider routing stays behind ManifoldGen.