MiniMaxMiniMax H3
Open-Weight Multimodal Video AI
MiniMax H3, also known as Hailuo 03, generates, references, and edits video in one model. Native 2K output, native stereo audio, and multi-reference control over character, camera, and voice.
MiniMax H3 is live on faktry: text-to-video, image-to-video, and reference-to-video, all with native stereo audio.
What Is MiniMax H3?
MiniMax H3 (Hailuo 03) is an open-weight, general-purpose multimodal video model from MiniMax. It unifies generation, referencing, and editing in a single model: one request can carry a character's identity from a photo, the camera language from a reference clip, and a voice from an audio recording. Every generation ships with native stereo audio, at resolutions up to 2K (1440p) and 24 fps.
MiniMax H3 at a glance
- Developer
- MiniMax
- Type
- Open-weight, general-purpose multimodal video model
- Resolution
- Up to 2K (1440p) at 24 fps
- Video length
- 5-15 seconds per generation
- Audio
- Native stereo sound on every generation
- Reference inputs
- Up to 9 images, 3 video clips, and 3 audio clips per request
- On faktry
- Live now: text-to-video, image-to-video, and reference-to-video
MiniMax H3, Live on faktry
Three ways to generate video with MiniMax H3, each with native stereo audio and billed per second of output.
Text to Video
Generate a video with native stereo audio directly from a text prompt.
Image to Video
Animate a starting image into video, with an optional end frame for first-to-last keyframe control.
Reference to Video
Generate video conditioned on multiple reference images, clips, and audio at once.
Duration, resolution, and aspect ratio are all configurable per generation.
What Makes MiniMax H3 Different
One model that generates, references, and edits video, instead of a separate tool for each job.
Native 2K with Stereo Audio
Every generation renders at up to 2K (1440p) resolution and 24 fps, with native stereo audio included by default.
Unified Multimodal Context
Text, images, video, and audio all enter the same context, so a single request can carry a subject's identity, a camera style, and a voice at once.
Deep Reference Input
Combine up to 9 reference images, 3 video clips (2-15s each), and 3 audio clips in a single generation, up to 12 files total.
Built-In Editing
Replace, remove, or add characters and objects, swap backgrounds, relight scenes, and adjust dialogue or vocal identity, with localized edits that keep the rest of the frame stable.
Legible Text Rendering
Renders readable title cards, signage, credits, and UI panels directly from a prompt.
Video-to-Video Motion Transfer
Carry the camera language and cutting rhythm from a reference clip into a new generation.
Capabilities
A closer look at what MiniMax H3 can generate on faktry.
Video Generation
See MiniMax H3 in Action
A look at the kind of video MiniMax H3 produces.
Where MiniMax H3 Fits In
One model covering the reference-heavy, edit-heavy workflows that used to need several separate tools.
Advertising & Branding
Produce on-brand video from reference images, brand voice audio, and a text prompt in a single generation.
E-Commerce & Product Visualization
Turn product photos into motion, with consistent framing, lighting, and legible on-screen text.
Gaming Cinematics
Generate cinematic shots with consistent characters and camera movement carried from a reference clip.
Film Titles & Animated Posters
Render title cards, credits, and animated poster art with legible, accurate text directly from a prompt.
How It Works
From prompt and references to a finished clip in three steps.
Add a prompt and references
Write a text prompt and optionally attach reference images, video clips, or audio for character, camera, or voice.
Generate
MiniMax H3 generates a clip up to 2K resolution with native stereo audio in one pass.
Refine or edit
Adjust the result with built-in editing: swap backgrounds, replace subjects, or tweak dialogue and voice.
Simple, Usage-Based Pricing
MiniMax H3 is billed per second of generated video, starting at 10 credits per second.
See Full PricingFrequently Asked Questions
What is MiniMax H3?
MiniMax H3 (Hailuo 03) is MiniMax's open-weight, general-purpose multimodal video model. It unifies text-to-video, image-to-video, reference-to-video, and video editing in a single model, with native stereo audio on every generation.
Is MiniMax H3 open-weight?
Yes, MiniMax released H3 with open weights. On faktry, it runs as a hosted endpoint through fal.ai, so there's no setup or self-hosting required.
How is MiniMax H3 different from other video models?
Most video pipelines need a separate model for generation, referencing, and editing. MiniMax H3 handles all three in one model, and accepts up to 9 images, 3 video clips, and 3 audio clips in a single request.
How many reference files can I use?
Up to 9 reference images, 3 reference video clips (2-15 seconds each), and 3 reference audio clips, for a maximum of 12 files per generation.
How much does MiniMax H3 cost on faktry?
Text-to-video and image-to-video are 10 credits per second of output. Reference-to-video is 12 credits per second. All generations range from 5 to 15 seconds.
Which other models can I use on faktry right now?
Flux 3, Sora 2, Veo 3.1, Kling 3.0, Seedance, and more are already live on faktry alongside MiniMax H3. Browse the full lineup on our AI Models page.