What Is MiniMax H4?

MiniMax H4 is a general-purpose multimodal video generation model developed by MiniMax. It can understand context across text, images, video, and audio and generate audiovisual video with native stereo sound. Current H3 workflows support 4–15 second output at 768P or up to 2K.

MiniMax H4 is not the same model as MiniMax M3. H3 is built for video generation and multimodal video workflows, while M3 belongs to MiniMax's language, coding, and agent model family.

Users sometimes search for the model as “MiniMaxH4” or “MiniMax-H3.” These are search variations of the standard model name MiniMax H4.

MiniMaxH4.org is an independent third-party platform that provides browser-based MiniMax H4 generation workflows, examples, prompts, pricing information, and practical model resources.

Looking for first-party destinations? See the official MiniMax H4 resources.

Quick answer

Inputs: text prompt, image references (first and/or last frame), video references, audio references (mode dependent).
Output: audiovisual video with native stereo audio, 4–15 seconds, 768p or 2K.
Cost: 6 credits per second at 768p, 12 credits per second at 2K.
Note: "MiniMax 4" is a common search spelling for the MiniMax H4 video model; it is not a separate product name used by this site.

Last verified: August 11, 2026

MiniMax H4 at a Glance

Field	MiniMax H4
Developer	MiniMax
Model type	General-purpose multimodal video generation model
Context inputs	Text, image, video and audio
Video duration	4–15 seconds
Output resolution	768P or up to 2K
Frame rate	24 FPS
Audio	Native stereo audiovisual generation
Base workflow families	FL2VA and Ref2VA
Local open-weight output	H3-Base at 768P
Full-system 2K stage	H3-Regenerate-2K

MiniMax H4 should be understood as a system rather than a single “2K checkpoint.” The open H3-Base generates the base audiovisual result, while the complete H3 workflow can use separate context-processing and 2K-regeneration stages.

Available on MiniMaxH4.org

Yes. MiniMaxH4.org provides access to supported MiniMax H4 workflows through its video generator. Credit costs are shown before each generation, and the exact availability of each mode is verified at generation time.

Sources

Facts above were verified against the following primary sources on 2026-08-11. Provider documentation can change.

How Is the MiniMax H4 System Structured?

The following components belong to MiniMax H3; the H4 architecture has not been published. The complete MiniMax H3 generation system has three major stages with different responsibilities.

H3-Context-IR

H3-Context-IR interprets complex multimodal instructions and relationships across text, images, video, and audio. It converts that context into an intermediate representation that H3-Base can use more reliably.

H3-Base

H3-Base performs the core audiovisual generation step and produces the base 768P result. The open-weight release includes FL2VA and Ref2VA task families.

H3-Regenerate-2K

H3-Regenerate-2K takes the base result together with the original context and regenerates the output at 2K. It is part of the complete H3 system rather than a separate user-facing video model.

Open H3-Base ≠ the complete hosted H3 system.

Open H3-Base vs the Full MiniMax H4 System

Open H3-Base versus the complete MiniMax H4 system
Capability	Open H3-Base	Complete H3 System
Core audiovisual generation	Yes	Yes
768P generation	Yes	Yes
FL2VA	Yes	Yes
Ref2VA	Yes	Yes
H3-Context-IR	Not included in the open-weight release	Available as part of the full workflow
H3-Regenerate-2K	Not part of H3-Base alone	Used for full 2K workflow
Local deployment	Yes	Hybrid/hosted components required for complete workflow

If you download MiniMax H4 locally, you are primarily deploying H3-Base. Reproducing the complete official 2K workflow requires the additional context-processing and regeneration stages. For setup details, see the MiniMax H4 Download guide and the MiniMax H4 ComfyUI Guide. For a source-linked explanation of which system components are public, read Is MiniMax H4 Open Source?.

Which MiniMax H4 Workflow Should You Use?

MiniMax H4 workflow selector
What You Have	Best Starting Workflow	Guide
Only a text prompt	Text-to-Video	MiniMax H4 Prompt Guide
One starting image	Image-to-Video	MiniMax H4 Image to Video
A first and last image	First + Last Frame	MiniMax H4 Image to Video
One existing video to transform or edit	Video-to-Video	MiniMax H4 Video to Video
Multiple images, videos, or audio references	Reference-to-Video	MiniMax H4 Reference Video
Local GPU / ComfyUI setup	Local H3-Base	MiniMax H4 ComfyUI
Model weights	FL2VA / Ref2VA	MiniMax H4 Download

Use Image-to-Video when an image must anchor the opening or ending frame. Use Video-to-Video when an existing source clip is the main object you want to transform or edit. Use Reference-to-Video when multiple assets each need separate roles such as identity, motion, camera, or voice.

MiniMax H4 FL2VA vs Ref2VA

MiniMax H4-Base is released in two task-specific workflow families. FL2VA covers text-to-video and first/last-frame audiovisual generation. Ref2VA handles multimodal reference generation using image, video, and audio references.

H3-Base workflow families FL2VA and Ref2VA
H3-Base Family	Best For
FL2VA	Text-to-video
FL2VA	First-frame image-to-video
FL2VA	Last-frame image-to-video
FL2VA	First + last frame
Ref2VA	Character reference
Ref2VA	Image reference
Ref2VA	Motion/video reference
Ref2VA	Camera reference
Ref2VA	Audio/voice reference

If you only need text or keyframe-based image-to-video, FL2VA is the relevant task family. If your workflow depends on multimodal identity, motion, camera or audio references, use Ref2VA. Both families are covered by the open Download MiniMax H4 weights.

MiniMax H4 vs MiniMax M3

MiniMax H4 and MiniMax M3 are separate model families.

MiniMax H4 versus MiniMax M3
	MiniMax H4	MiniMax M3
Primary purpose	Video generation	Language, coding and agent tasks
Main output	Audiovisual video	Text, code and agent/tool actions
Relevant on MiniMaxH4.org	Yes	No

If your goal is AI video generation, MiniMax H4 is the relevant model. MiniMax M3 is not the video-generation version of H3. See the dedicated MiniMax M3 vs H3 comparison for the full breakdown.

Try MiniMax H4 Online

MiniMaxH4.org provides browser-based access to supported MiniMax H4 generation workflows. Choose the available input mode, resolution, and duration in the generator and review the credit quote before submitting.

Try MiniMax H4 See MiniMax H4 Pricing MiniMax H4 API10 Free Credits Daily · Up to 30 Free Credits

MiniMax H4 FAQ

What is MiniMax H4?

MiniMax H4 is MiniMax's general-purpose multimodal video generation model. It understands context across text, images, video, and audio and generates audiovisual video with native stereo sound. Current H3 workflows support 4–15 second output at 768P or up to 2K resolution, and the open-weight release covers the FL2VA and Ref2VA task families.

Is MiniMaxH4 the same as MiniMax H4?

Yes. “MiniMaxH4” is a common no-space search variation of MiniMax H4. Both searches refer to the same general-purpose multimodal video generation model from MiniMax. MiniMaxH4.org uses “MiniMax H4” when referring to the model because that is the standard model name; spellings like “MiniMax-H3” are the same entity.

Is MiniMax H4 the same as MiniMax M3?

No. MiniMax H4 is a video generation model built for multimodal video workflows, while MiniMax M3 belongs to MiniMax's language, coding, and agent model family. They serve different purposes, and only H3 is relevant to video generation on MiniMaxH4.org. If your goal is AI video, H3 is the model to use.

How long can MiniMax H4 videos be?

Current MiniMax H4 workflows support output durations from 4 to 15 seconds. The exact duration available depends on the generation mode and provider. Lower-resolution drafts and longer clips each carry different credit costs, so the generator shows the estimated credit requirement before you submit.

Does MiniMax H4 support 2K?

Yes. The complete MiniMax H4 system supports output up to 2K. The open H3-Base generates the base 768P result, while the H3-Regenerate-2K stage is used in the complete 2K workflow. On MiniMaxH4.org, 768p and 2K options are available for supported H3 workflows.

Does MiniMax H4 generate audio?

Yes. MiniMax H4 is an audiovisual generation model and supports native stereo audio generation. Audio availability can depend on the selected generation mode and provider, so check the generator settings before starting a task when sound matters to the result.

Does MiniMax H4 support image-to-video?

Yes. The FL2VA task family supports first-frame, last-frame, and first-plus-last-frame image-to-video generation using supplied keyframes as composition anchors. See the MiniMax H4 Image to Video guide for input requirements, first/last-frame control, prompting, and workflow selection when starting from a still image.

Can MiniMax H4 use video and audio references?

Yes. Ref2VA supports multimodal reference workflows using images, videos, and audio, subject to the current input limits. Reference inputs can control identity, motion, camera, or voice in the generated clip. The available reference modes are shown in the generator before you submit.

Can MiniMax H4 run locally?

Yes. The open H3-Base weights can be deployed locally, including the FL2VA and Ref2VA task families. The complete official 2K workflow also uses the additional H3-Context-IR and H3-Regenerate-2K stages, so local deployment primarily reproduces the base 768P result. Setup details are covered in the download and ComfyUI guides.

Setup guides

Step-by-step setup instructions and workflow guidance for running MiniMax H4.

MiniMax H4 APIEndpoints, pricing, Python, cURL and input limits.MiniMax H4 Reference VideoCombine image, video, and audio references when each asset needs a different control role.MiniMax H4 Image-to-VideoAnimate a supplied image when the opening or ending composition needs to stay anchored.MiniMax H4 Video to VideoTransform, edit, continue, or transfer motion from an existing source video.MiniMax H4 ComfyUI GuideRun H3 locally in ComfyUI with T2V, I2V and R2V.MiniMax H4 DirectorCompare community timeline nodes, GitHub projects and Director-style H3 workflows.MiniMax H4 Prompt GuideNative prompt structure for T2V, I2V, FL2V and R2V.Download MiniMax H4FL2VA and Ref2VA open weights from Hugging Face.MiniMax H4 TurboRun H3 faster with Turbo LoRA, LightX2V, and step-reduced samplers.MiniMax H4 LoRACurated LoRA archetype directory — Turbo, style, character, motion, and camera.MiniMax H4 vs Seedance 2.0Specs, workflows and a complex-motion stress test.