HopBase
MiniMax's new open general-purpose multimodal video model, Hailuo H3, is now live on HopBase through an official direct connection. Sign up and start calling it self-serve. The cover of this post is the untouched first frame of a real H3 output (2K text-to-video).
What it does
- Three workflows: text-to-video, image-to-video (first / last frame anchoring), and multimodal reference — images, video clips and audio jointly steer subject, motion and rhythm.
- 4-15 second clips at 768P or 2K, aspect ratios from 21:9 to 9:16, 24 FPS.
- Native stereo audio on every clip, with in-model dialogue and voice-over via TTS covering 11 languages — no separate dubbing pass.
- Billed only on completed clips: failed or safety-filtered generations are never charged.
Pricing: 85% of official list
| Item | Official list | HopBase standard |
|---|---|---|
| 768P output (per second) | ¥0.50 | ¥0.425 |
| 2K output (per second) | ¥0.80 | ¥0.68 |
| Input images (each beyond the first 5) | ¥0.20 | ¥0.17 |
Reference-video input is billed per input second at the output tier; reference audio is free. Full details on the pricing page.
Three-minute quickstart
curl https://api.hop-base.com/v1/video/generate \
-H "Authorization: Bearer sk-your-key" \
-H "Content-Type: application/json" \
-d '{
"model": "MiniMax-H3",
"content": [{"type": "text", "text": "Aerial drone shot over snow mountains above a sea of clouds"}],
"resolution": "2K", "duration": 4, "ratio": "16:9"
}'The submission returns a task ID; poll GET /v1/video/tasks/{task_id} until completed to get a stable download URL. The request body mirrors MiniMax's official v2 contract, so existing official-SDK code migrates with zero changes. Full contract in the MiniMax H3 docs.