MiniMax H3 API

Resolution tiers and pricing

Both tiers carry native synchronized audio; resolution is the only difference. 768P saves a third for drafts, feeds and small players; native 2K ships without any upscaling step. Charged only on success, refunded on failure.

Aliases: hailuo-h3 and minimax-h3 point at the same model — Hailuo is the product name of the MiniMax video line.

Input asset pricing

Reference assets are priced separately, on top of the output seconds — a reference video bills by its own length, reference audio is free, and the first 5 reference images are free.

Example: a 10-second 2K output with a 7-second reference video → 10 × $0.145 + 7 × $0.135 = $2.395. Another: a 5-second 2K output with 8 reference images → 5 × $0.145 + 3 × $0.054 = $0.887.

A reference video longer than 15 seconds is rejected at create time: we probe each reference clip’s real duration and return a clear 400 before any task is created or credits frozen.

Authentication

All requests carry the API key in the header:

Authorization: Bearer YOUR_API_KEY

Video Generation

# Text-to-video (model: minimax-h3 — the alias hailuo-h3 also works)
curl -X POST https://api.apimodels.app/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3",
    "prompt": "a lighthouse keeper climbs a spiral staircase at dusk, warm lamplight, waves crashing outside",
    "resolution": "768p",
    "ratio": "16:9",
    "duration": 10
  }'

# Poll until completed (result files are kept 7 days)
curl "https://api.apimodels.app/v1/video/generations?task_id=TASK_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"

# All reference inputs are optional — prompt alone is text-to-video.
# Attach images / video_list / audio_list to steer character, motion and voice
# in the same request (see the multimodal example in the Python / Node tabs).

Parameters

Billing

total = duration × per-second price
      + reference video seconds × input-video price
      + max(0, images − 5) × $0.054

• Per-second price: $0.145 at 2K, $0.097 at 768P; input-video price: $0.135/s at 2K, $0.081/s at 768P; reference audio is free.
• Charged only on success: a failed generation is fully refunded automatically, input-asset charges included.

Notes

• Async: create returns a taskId; poll GET /v1/video/generations?task_id= until completed, or pass callback_url and we call you back on completion. Result files are deleted after 7 days — download or re-host promptly.
• Every reference input is optional and they combine freely — images only, audio only, or images + video + audio together. Nothing attached means plain text-to-video.
• Parameter errors (duration out of range, ratio outside the enum, too many assets) return a synchronous 400 at create time — no task is created and no credits are frozen.
• Prompts and outputs go through content moderation; a moderation rejection is treated as a failure and refunded.

Try it in the Playground