H3 Max API: Endpoints, Parameters & Price per Second
H3 Max is available through three APIs: fal.ai, MiniMax's own platform and OpenRouter. All three charge per second of output, starting at $0.05 per second at 480P. Only fal offers 1080P and the half-price H3 Max Turbo. A 5-second 768P clip costs $0.40 on any of them at list price.
If you just want videos without writing code, you can use the H3 Max video generator on this site instead.
H3 Max API pricing by provider
List prices per second of generated video, as published by each provider:
| Provider | Model ID | 480P | 768P | 1080P |
|---|---|---|---|---|
| fal.ai | minimax/h3-max/… | $0.05 | $0.08 | $0.16 |
| fal.ai (Turbo) | minimax/h3-max-turbo/… | $0.025 | $0.04 | $0.08 |
| MiniMax | MiniMax-H3-Max | $0.05 | $0.08 | — |
| OpenRouter | minimax/hailuo-3-max | from $0.05 | up to $0.08 | — |
Notes:
- fal ran a 50% launch discount on H3 Max that ended on September 30, 2026. The table shows the regular prices.
- MiniMax also charges for some inputs: the first two images are free, then $0.074 per extra image; audio input is free.
- fal's 1080P is a latent refinement of a native 768P generation, not a separate native resolution.
Cost per clip (fal list price)
| Clip | H3 Max 480P | H3 Max 768P | H3 Max 1080P | Turbo 480P | Turbo 768P | Turbo 1080P |
|---|---|---|---|---|---|---|
| 5 s | $0.25 | $0.40 | $0.80 | $0.125 | $0.20 | $0.40 |
| 10 s | $0.50 | $0.80 | $1.60 | $0.25 | $0.40 | $0.80 |
| 15 s | $0.75 | $1.20 | $2.40 | $0.375 | $0.60 | $1.20 |
fal.ai endpoints
| Endpoint | Use |
|---|---|
minimax/h3-max/text-to-video | Text to video |
minimax/h3-max/image-to-video | First frame, last frame or both |
minimax/h3-max-turbo/text-to-video | Text to video, half price |
minimax/h3-max-turbo/image-to-video | Image to video, half price |
Turbo takes the same inputs. fal describes it as a faster, distilled version of H3 Max that is still in preview; see H3 Max Turbo for how the two compare.
Request parameters (fal)
| Parameter | Type | Notes |
|---|---|---|
prompt | string, required | The video description |
prompt_expansion_mode | string, required | disabled, balanced (default, about 1 s) or quality (up to ~30 s of rewriting) |
duration | integer | 5–15 seconds, default 5 |
resolution | enum | 480P, 768P (default) or 1080P |
aspect_ratio | enum | Text to video only: 21:9, 16:9 (default), 4:3, 1:1, 3:4, 9:16 |
image_url | string | Image to video: first frame. The output follows this image's shape |
end_image_url | string | Image to video: last frame. Can be sent alone for end-only keyframes |
target_audio_url | string | Audio of at least 2 seconds (max 15 MB) that replaces the generated soundtrack |
seed | integer | Fixed seed for repeatable results |
enable_safety_checker | boolean | Default true |
sync_mode | boolean | Return the video as base64 instead of a URL |
The response contains video.url, plus expanded_prompt (the rewritten prompt sent to the model, when expansion changed it) and timings.
A minimal request
curl --request POST \
--url https://fal.run/minimax/h3-max/text-to-video \
--header "Authorization: Key $FAL_KEY" \
--header "Content-Type: application/json" \
--data '{
"prompt": "A paper boat drifts down a rain-filled street gutter, low tracking shot, soft rain and distant thunder.",
"prompt_expansion_mode": "balanced",
"duration": 5,
"resolution": "480P",
"aspect_ratio": "9:16"
}'For production use, submit jobs through fal's queue API or client libraries and poll for the result, rather than holding a synchronous request open. How you write the prompt matters as much as the parameters — see the H3 Max prompt guide.
Which provider should you use?
- fal.ai if you need 1080P, the cheaper Turbo model, or a single API for many video models.
- MiniMax if you want to buy directly from the model's developer. It lists 480P and 768P for H3 Max, plus the base H3 model with 2K output.
- OpenRouter if you already route other model calls through it. Its H3 Max listing supports 480p and 768p, clips up to 15 seconds and six aspect ratios.
The base model, MiniMax H3, is also available as open weights you can run yourself — see our MiniMax H3 ComfyUI guide, and check the license's territory limits first.
FAQ
Is there an official H3 Max API?
Yes. MiniMax lists MiniMax-H3-Max on its own API platform. fal.ai, which released H3 Max jointly with MiniMax, offers it too.
Does the H3 Max API support 1080P? On fal.ai, yes: 1080P at $0.16 per second, refined from a 768P generation. MiniMax's and OpenRouter's listings stop at 768P.
How much does H3 Max Turbo cost? On fal.ai, exactly half of H3 Max: $0.025, $0.04 and $0.08 per second at 480P, 768P and 1080P. MiniMax's price list does not include Turbo.
How long can an H3 Max video be? 5 to 15 seconds per request on fal.ai. OpenRouter lists a 15-second maximum as well.
Can I use H3 Max without an API key? Yes. On h3maxvideo.net you sign in with Google, get free credits for a first 480P video and pay with credit packs from $2.99 — see pricing.
Sources: fal.ai H3 Max text to video and image to video API pages, MiniMax pay-as-you-go pricing, OpenRouter: minimax/hailuo-3-max. Prices change often; check the provider before you budget. Last updated: October 1, 2026.
h3maxvideo.net is independent and not affiliated with MiniMax, fal.ai or OpenRouter.