Ready-to-Use H3 Workflow Templates
Start with ComfyUI templates for H3 text-to-video, image-to-video, first-and-last-frame, and reference-to-video generation, then customize nodes and parameters for your project.
MiniMax H3 in ComfyUIBuild Custom Multimodal AI Video Workflows
Use MiniMax H3 and ComfyUI to connect prompts, models, parameters, images, video, audio, and outputs in reusable node-based workflows. Create text-to-video, image-to-video, first-and-last-frame, and reference-driven video with synchronized visuals, motion, and sound.
Why Use MiniMax H3 with ComfyUI?
ComfyUI turns MiniMax H3 into a customizable production pipeline, letting you choose how prompts, references, preprocessing, generation parameters, audio, and outputs connect instead of relying on a fixed interface.
Multimodal Creative Control
Combine text, images, video, and audio references to guide character identity, visual style, motion, camera behavior, voice, ambience, effects, and music inside one workflow.
Joint Video & Native Audio Generation
Generate visuals, dialogue, environmental sound, effects, and music as part of the same H3 pipeline, reducing the need to assemble every audio layer manually afterward.
Reference-to-Video Workflows
Assign clear roles to creative assets: use images for identity or style, video for motion and camera language, and audio for voice or sound direction.
Local or API-Based Generation
Run open-weight H3 models locally for deeper model and node control, or use hosted API workflows when you want faster setup without downloading large models or managing local inference.
Reusable Production Pipelines
Save complete workflows, reuse parameters and reference routing, add preprocessing or post-processing nodes, and extend fast experiments toward higher-resolution production outputs.
MiniMax H3 ComfyUI Workflow Examples
Explore how different H3 workflows organize prompts, frames, references, motion, camera direction, and native audio for reusable AI video production.
A confident performer executes a fast dance routine on a neon-lit stage. The camera tracks vertically with the movement before making a smooth arc around the subject. Preserve natural body mechanics and facial expression. Add synchronized footsteps, crowd ambience, rhythmic electronic music, and cinematic lighting with clear subject separation.
Animate the starting image into a coordinated group performance. Preserve the identity, clothing, relative position, and proportions of all five performers. Add synchronized choreography, clear limb movement, a sweeping camera move, stage ambience, and music while keeping the original visual style stable.
Begin from the supplied first frame and finish on the supplied last frame. Create a graceful performance between them with controlled torso movement, expressive arm gestures, consistent identity, and a smooth camera transition. Make the motion naturally arrive at the final pose and synchronize the rhythm with cinematic music.
Use Image 1 for product identity and geometry, Video 1 for the disassembly motion, and Audio 1 for timing. Show the product separating into clearly defined components before reassembling. Preserve branding, shape, materials, and proportions, and finish with a stable presentation-ready hero shot.
Use Video 1 only as the movement and camera reference. Recreate its accelerating backward scenery, rhythmic frame-drop effect, and camera energy in a new cinematic environment. Preserve the generated scene structure while adding controlled visual trails, synchronized impacts, and dynamic editing rhythm.
Generate a person performing push-ups with believable weight, muscle tension, breathing, and visible effort. Keep body proportions stable, facial emotion natural, and framing steady. Generate synchronized breathing, clothing movement, floor contact sounds, subtle room ambience, and restrained motivational music with the video.
Run MiniMax H3 Locally or Through an API Workflow
ComfyUI supports two practical H3 approaches. Local workflows offer deeper control over models and nodes, while API workflows reduce setup and hardware requirements by running generation on hosted infrastructure.
| Workflow Consideration | Local H3 Workflow | H3 API Workflow |
|---|---|---|
| Setup | Download the required H3 model files and place them in the correct ComfyUI directories | Connect a hosted H3 endpoint without downloading the full local model stack |
| Hardware | Requires sufficient GPU memory, storage, and system resources for the selected model configuration | Generation runs remotely, reducing local GPU and storage requirements |
| Model & Node Control | Greater control over model files, parameters, custom nodes, preprocessing, and output management | Workflow control remains in ComfyUI, while available model settings depend on the hosted endpoint |
| Maintenance | You manage model downloads, updates, dependencies, and hardware compatibility | The provider manages inference infrastructure and model deployment |
| Scaling | Best for controlled personal pipelines and reusable workflows within available hardware limits | Better suited to rapid onboarding, shared teams, and production workloads that need flexible capacity |
| Best For | Advanced ComfyUI users who want maximum pipeline ownership and customization | Creators who want faster setup without managing large models or high-end local GPUs |
How to Use MiniMax H3 in ComfyUI
Set up H3, choose the right workflow template, connect your creative inputs, and generate a reusable video-and-audio pipeline in four steps.
- 01
Update ComfyUI & Find H3 Templates
Use a recent ComfyUI version, open the Template Library, and search for MiniMax H3 workflows such as H3 T2V, H3 I2V, or H3 Reference-to-Video.
- 02
Install Models or Configure the API
For local use, download the diffusion model, text encoder, Video VAE, Audio VAE, and any workflow-specific files into the correct ComfyUI directories. For hosted use, configure the H3 API nodes instead.
- 03
Choose the Workflow & Connect Inputs
Use Text-to-Video for prompts, Image-to-Video for a still image, First-and-Last-Frame for controlled transitions, or Reference-to-Video when combining image, video, and audio guidance.
- 04
Write the Prompt, Generate & Iterate
Describe the scene, subject, action, camera, style, dialogue, and sound. Set the generation parameters, run the graph, review the video and audio outputs, then save the workflow for reuse.
Frequently Asked Questions
Is MiniMax H3 supported in ComfyUI?
Yes. MiniMax H3 can be used through ComfyUI workflows for Text-to-Video, Image-to-Video, first-and-last-frame generation, and multimodal Reference-to-Video. Start from an available H3 workflow template and customize the graph for your project.
Do I need custom nodes to use MiniMax H3 in ComfyUI?
Not necessarily. Standard H3 workflows can begin from supported ComfyUI templates, so you do not need to build an entire custom-node pipeline from scratch. Additional nodes are useful when you want custom preprocessing, routing, automation, or post-processing.
Can I run MiniMax H3 locally in ComfyUI?
Yes. Open-weight H3 models can run in local ComfyUI workflows, giving you control over model files, parameters, nodes, references, and outputs. Hardware requirements depend on the selected model, resolution, duration, and workflow configuration.
Can I use H3 in ComfyUI without downloading the model locally?
Yes. An API-based workflow can use ComfyUI as the node interface while H3 generation runs on hosted infrastructure. This reduces local setup, storage, and GPU requirements, although available parameters depend on the API provider.
Does MiniMax H3 support Image-to-Video in ComfyUI?
Yes. Use an image as the starting frame, then add a prompt to direct movement, camera behavior, scene development, visual continuity, and sound. The workflow can preserve the main subject, style, and composition while animating the scene.
Can H3 generate audio together with video?
Yes. H3 supports joint audiovisual generation, allowing dialogue, ambience, sound effects, music, and other audio elements to be produced with the visual content in the same workflow.
Can I use images, videos, and audio as references?
Yes. Reference-to-Video workflows can combine multiple media types. Clearly assign each asset a role in the prompt—for example, an image for character identity, a video for movement or camera language, and audio for voice or sound direction.
What is the difference between H3 Text-to-Video and Reference-to-Video?
Text-to-Video follows a written prompt as its primary instruction. Reference-to-Video adds image, video, or audio guidance, giving you more precise control over identity, style, motion, camera behavior, voice, and sound.
Do I need a powerful GPU for H3 ComfyUI workflows?
Local generation requires suitable GPU memory and system resources, and performance varies by model configuration, resolution, duration, and node graph. If local inference is impractical, use an API workflow to run generation remotely.
How should I write MiniMax H3 prompts in ComfyUI?
Use a structured prompt covering scene, subject, action, camera, visual style, dialogue, and sound. For multimodal workflows, also state what each reference controls instead of asking H3 to interpret every asset automatically.
What is MiniMax Design, and will it support ComfyUI?
MiniMax Design is an Agent-powered creative platform designed to help plan tasks, organize assets, select models, and coordinate production workflows. ComfyUI integration has been announced as coming soon, so availability may depend on the current product release.