H3 Fast

The fastest, least expensive video model; 480p, 5-15 seconds, with audio. For drafts and bulk work.

modellane/h3-fast · Text to video · MiniMax

The fastest and least expensive variant in the H3 family. It produces 5-15 second, 24fps videos with audio at 480p. It is tuned for high-volume draft production, bulk variation sweeps, and quick tests at the idea stage. An ideal first step: turn an idea into motion within seconds, then finish with H3 Max or Seedance 2.5.

Vendor: MiniMax
Task: Text to video
Output: Video
Price: $0.055 · per second
Longest run: 15 seconds

Input parameters

Prompt prompt — required, at most 7000 characters Describe what you want produced as concretely as you can.
Duration duration — optional, default: 5, range: 5–15 sn, affects price A longer duration raises the price directly.
Aspect ratio ratio — optional, default: 16:9, options: 16:9, 9:16, 1:1, adaptive "adaptive" picks the ratio from the source.
Prompt expansion prompt_expansion — optional, advanced, default: false When on, the model enriches the prompt with its own rules.

Recommended use cases

Bulk draft production — Produce dozens of variations on an idea within seconds and cut them down.
Idea-stage tests — Test the scene in motion before committing to production.
Budget-friendly content — Produce high-volume social media content at a low unit cost.
First pass, then upgrade — Draft with H3 Fast and finish with H3 Max or Seedance 2.5.

Where it is strong

The video model with the lowest per-second cost
Very fast generation, an instant feedback loop
24fps and synchronised audio as standard
Four aspect ratios: horizontal, vertical, square, adaptive

Limits

480p only — not suitable for broadcast quality
15-second ceiling
No resolution choice; always 480p

Example prompt

A cat stretching lazily on a sunlit windowsill, tail curling, then curling back into a ball, warm afternoon light, cozy interior, gentle movement

Example request

POST /v1/queue/modellane/h3-fast

Playground

The identifier, price, parameters, and example requests on this page can be read without signing in. Running the model from the panel sends a real generation request and charges the project balance; the playground therefore requires an account.