Try “Veo”, “voiceover”, “thumbnail” or “Descript” · Esc to close

MiniMax H3 (Hailuo 3.0) 🧠 Model Freemium

MiniMax's open-weight video model: native 2K, stereo sound, 15-second clips

Video models · Free plan, paid from $14.99/mo ★ 0 · MiniMax H3 Community License · archived

8.2editor score
Visit MiniMax H3 (Hailuo 3.0) ↗
GitHub stars
0
Stars this week
–
Forks
0
Licence
MiniMax H3 Community License
Last push
–
Maintainer
MiniMax

Where you can use MiniMax H3 (Hailuo 3.0)

Apps in this directory that let you generate with MiniMax H3, from their own documentation.

See every app × model on the model map →

Good for

About MiniMax H3 (Hailuo 3.0)

What it is

Hailuo is MiniMax's video model line. It powers the Hailuo AI app. The current generation, MiniMax H3 (also called Hailuo 3.0), launched on 31 July 2026. It takes text, images, video and audio as input and returns video with stereo sound generated in the same pass. Unlike the earlier Hailuo 02 and 2.3, H3 is also published as open weights on Hugging Face. Some parts stay API-only, including the 2K regeneration stage.

Key features

  • Native 2K output (2560x1440, 24 fps), with a cheaper 768p mode
  • Clips up to 15 seconds with stereo dialogue, ambience and music
  • Up to 9 reference images, 3 video clips and 3 audio clips per request
  • Instruction-based editing that changes one element and leaves the rest of the shot alone
  • Open weights (33B parameters) under the MiniMax H3 Community License

Where you can use it

The Hailuo AI app and the MiniMax API. H3 or Hailuo 2.3 is documented in Krea, Leonardo.Ai, Luma's app, Higgsfield, Freepik, invideo AI and ComfyUI partner nodes. OpenArt lists Hailuo 02. The open weights (MiniMaxAI/MiniMax-H3) need a multi-GPU server to run.

Pricing and rights

The Hailuo AI app's free users get watermarked downloads. Standard costs $14.99 a month for 1,000 credits, and plans go up to Max at $199.99. Hailuo's subscription terms say paid users keep the rights to their output, including commercial use, and get watermark-free downloads. MiniMax says H3's API costs a fraction of competitors' prices per second, but we did not find a published rate. The open-weight licence has an application form for the USA, EU, UK and South Korea.

Who it is for

Anime and stylised creators, short-form video makers, and studios that want an open-weight model they can host themselves.

Verdict

H3 moves Hailuo from a budget favourite to a front-line model with sound and open weights. Self-hosting needs serious hardware, and the licence is not a standard open-source licence.

text-to-video image-to-video open-weights native-audio minimax

Pros

  • Native 2K with stereo audio
  • Open weights for the 2026 flagship
  • Strong at anime and stylised motion
  • Paid plans grant commercial rights

Cons

  • Free downloads are watermarked
  • Self-hosting needs about 4 datacentre GPUs
  • Custom community licence with regional application form
  • No published per-second API price found

Similar AI models

All video models →

Veo 3.1 🧠 ModelPaid

Google DeepMind's text- and image-to-video model with native sound, up to 4K

Google DeepMind · 2025-10

8.8 Visit ↗

Kling 3.0 🧠 ModelFreemium

Kuaishou's video model with native multilingual audio, lip sync and 15-second clips

Kuaishou · 2026-02

8.6 Visit ↗

Seedance 2.5 🧠 ModelPaid

ByteDance's video model: 30-second clips with audio, many references and local edits

ByteDance Seed · 2026-07

8.6 Visit ↗

Runway Gen-4.5 🧠 ModelFreemium

Runway's own video model, now with native audio and one-minute multi-shot scenes

Runway · 2025-12

8.2 Visit ↗

Wan 3.0 🧠 ModelOpen source

Alibaba's video family: open-weight Wan 2.x and the 30-second Wan 3.0 API

Alibaba (Tongyi Wan) · 2026-08 · open weights

8.0 Visit ↗

LTX-2.5 🧠 ModelOpen source

Lightricks' open-weight video model with synchronised audio and 4K output

Lightricks · 2026-08 · open weights

7.8 Visit ↗

More from MiniMax

MiniMax MCP 🔌 MCP serverFreemium

Official MiniMax server for TTS, voice cloning, Hailuo video and image generation

★ 1.6k

Popular searches