How to Use MiniMax H3 Max API

Use the current provider documentation before sending a production request.

fal.ai request

Step 1: Sign Up for fal.ai

fal.ai is currently the most prominent third-party provider hosting MiniMax H3 and H3 Max. Go to fal.ai and create an account — new users get a small one-time credit allocation that's enough for several test generations.

【FILL: screenshot of fal.ai signup page】

Step 2: Get Your API Key

Once logged in, go to your fal.ai dashboard → Settings → API Keys. Create a new key and copy it. You'll use this key to authenticate every request.

# Set your key as an environment variable (don't hardcode it in scripts)
export FAL_KEY="your-api-key-here"

You can also use OpenRouter as an alternative provider — the API pattern is similar but you'll use an OpenRouter key instead. See OpenRouter's documentation for their H3 Max model path and pricing.

Step 3: Make Your First Request

Install the fal Python client and send your first text-to-video request:

Step 4: Common Issues and Fixes

Here are the most common errors new H3 Max users encounter and how to solve them:

"Model not found" or "endpoint not available" — The model path may have changed. Always check fal.ai/models for the current H3 Max model ID. As of September 2026, the path includes "minimax/h3-max" but this can change.
"Insufficient credits" — Your free allocation ran out. Check your billing page on fal.ai and add a payment method. A single 5-second H3 Max generation at 480p costs approximately $0.03 on the promo rate (ended September 7, 2026). See our pricing page for post-promo rates.
"Request timeout" or "generation took too long" — H3 Max should return in under 3 seconds for a 5-second clip at 480p. If it takes longer, the provider may be under load. Try again or switch resolution to 480p if you requested 768p.
"Video has no audio" or "audio is out of sync" — Make sure you're calling the H3 Max endpoint (not the base H3). Audio sync is an H3 Max feature. If audio is still missing, check the response object — some providers return audio as a separate URL that needs to be merged client-side.
"Resolution not supported" — H3 Max supports 480p and 768p. Requesting 1080p will fail. If you need higher resolution, use the base H3 model instead.

import os, requests
response = requests.post(
    "https://fal.run/minimax/h3-max/text-to-video",
    headers={"Authorization": f"Key {os.environ['FAL_KEY']}"},
    json={"prompt": "Your prompt", "prompt_expansion_mode": "balanced"},
)
response.raise_for_status()
video_url = response.json()["video"]["url"]
print(video_url)

Frequently Asked Questions

Is MiniMax H3 Max on fal.ai?

Yes. fal.ai currently hosts H3 Max and it is the most prominent third-party provider. You can access it at fal.run/minimax/h3-max/text-to-video (verify the exact model path in fal's current documentation).

How do I get an H3 Max API key?

Sign up at fal.ai, generate an API key from your dashboard, and use it in the Authorization header of your requests. The same key works across all fal-hosted models. Check OpenRouter and the official MiniMax platform for alternative access options.

How long does H3 Max take to generate 5 seconds of video?

MiniMax states under 3 seconds for a 5-second clip. Actual wall-clock time will vary based on provider queue depth, your geographic location relative to the inference endpoint, and current demand.

How much does one H3 Max API call cost?

It depends on the resolution and the provider's current rate. See our /pricing/ page for the latest breakdown. At post-promo estimated rates, a single 5-second 480p generation costs roughly $0.125.

Last updated: September 10, 2026