ComfyMax v0.4.0 — FlashVSR, Prompt Library & LM Studio Improvements

ComfyMax v0.4.0

ComfyMax v0.4.0 brings the latest tested development changes into the public release.

Highlights

• Added integrated FlashVSR support
• Added the Prompt Library for storing and reusing approved prompts
• Improved LM Studio integration and model handling
• Fixed Send to ComfyUI after editing a generated prompt
• Improved setup and startup scripts
• Added FlashVSR installation and validation tools
• Added additional tests for FlashVSR, LM Studio, and Prompt Library functionality
• Updated documentation and workflow instructions

This release has been tested through the ComfyMax development branch before being merged into the public version.

For new installations, follow the instructions in INSTALLATION.md.

ComfyMax v0.3.0

ComfyMax v0.3.0

ComfyMax v0.3.0 focuses on making the application easier to extend, maintain and use with custom ComfyUI workflows.

⭐ New: Workflow Mapper

The biggest addition in v0.3.0 is the new Workflow Mapper.

You can now export a workflow from ComfyUI using Save (API Format), upload it to the Workflow Mapper and let ComfyMax analyze the workflow automatically.

The Workflow Mapper can:

• detect workflow nodes and editable inputs;
• automatically suggest mappings for prompt, duration, resolution, aspect ratio, steps and seed;
• detect reference-image inputs;
• suggest MiniMax H3 model mappings;
• show mapping confidence and compatibility information;
• validate node IDs and inputs;
• detect possible duplicate mappings;
• generate the required ComfyMax mapping JSON;
• install the workflow and mapping directly into ComfyMax.

This makes it much easier to experiment with and add your own ComfyUI workflows without manually creating mapping files.

🔄 New: ComfyMax Updater

A new Update_ComfyMax.bat makes updating a Git installation easier.

The updater:

• checks GitHub for new versions;
• protects locally modified tracked files;
• leaves untracked user files untouched;
• downloads the latest ComfyMax changes;
• updates Python dependencies when required.

📖 Improved Documentation

Installation instructions have been moved to a separate:

INSTALLATION.md

This provides new users with a shorter step-by-step installation guide without having to search through the full README.

The README has also been updated with information about the Workflow Mapper, updater and current ComfyMax features.

🎬 Existing v0.2 Features

ComfyMax continues to include:

• MiniMax H3 Ref2VA workflow support;
• up to 9 reference images;
• LM Studio prompt generation;
• prompt review and approval before rendering;
• Scene Builder with multi-turn dialogue;
• Video Gallery;
• ComfyUI model selection and settings;
• GPU and VRAM monitoring;
• manual ComfyUI and LM Studio model unloading;
• LAN access from tablets and other devices.

⚠️ Before Using a Workflow

Always test the workflow directly in ComfyUI first.

The supplied example workflow in workflow_example can be used to verify that the required nodes and models are installed correctly.

For custom workflows, export them from ComfyUI using Save (API Format) before importing them into the Workflow Mapper.

---

Thanks to everyone testing ComfyMax and providing feedback.

ComfyMax is still evolving, and real-world feedback is very welcome.

ComfyMax v0.2.0

ComfyMax v0.2.0

This release significantly expands ComfyMax beyond the original prompt-and-render workflow.

What's New

🎬 Scene Builder

• Build scenes step by step before generating the final H3 prompt.
• Define characters, locations, actions, camera style, lighting and atmosphere.
• Create multi-turn dialogue between characters.
• Add actions performed while characters are speaking.
• Review and edit the complete scene before sending it to the main prompt generator.
• Scene Builder questions are stored in a separate configuration file for easy customization.

🎞️ Video Gallery

• Browse generated ComfyUI videos directly inside ComfyMax.
• Search, filter and sort generated videos.
• Choose between multiple gallery layouts.
• Play videos directly in the interface.
• Download or delete generated videos.
• Open videos directly in Windows Explorer.
• Scan the configured ComfyUI output folder and its subfolders.

🖼️ Extended Reference Image Support

• Added workflows and mappings supporting up to 9 reference images.
• Reference images are passed to compatible multimodal LM Studio models in the correct order.

🧠 LM Studio Improvements

• Improved LM Studio integration.
• Added an unload option to free VRAM when LM Studio has been used for other purposes.
• Better VRAM management between LM Studio and ComfyUI.

⚙️ Settings Improvements

• Configure and save the ComfyUI output folder.
• Test the configured output folder directly from ComfyMax.
• Improved management of ComfyUI and LM Studio settings.

📊 GPU Monitoring

• Live NVIDIA GPU monitoring.
• Monitor GPU load, VRAM usage, temperature and power during prompt generation and rendering.

🔧 Workflow Improvements

• Improved workflow mappings.
• Better handling of MiniMax H3 Ref2VA workflows.
• Review and explicitly approve generated prompts before rendering.
• Finished H3 prompts can also be pasted directly into ComfyMax without using LM Studio.

Before Using ComfyMax

Please test the supplied workflow in the workflow_example folder directly in ComfyUI first.

This confirms that all required nodes and models are installed and that the workflow works correctly before ComfyMax is added to the pipeline.

ComfyMax remains under active development. Feedback, bug reports and suggestions are very welcome.

ComfyMax v0.1.0

ComfyMax v0.1.0

First public release of ComfyMax.

ComfyMax is a lightweight Streamlit frontend designed to make MiniMax H3 video generation with ComfyUI and LM Studio easier to use.

Highlights

• MiniMax H3 Ref2VA workflows with 1, 2 or 3 reference images
• LM Studio integration for multimodal H3 prompt generation
• Review and edit generated prompts before rendering
• Manual H3 prompt entry without LM Studio
• Explicit prompt approval before sending to ComfyUI
• Duration, resolution, aspect ratio, steps and seed controls
• Configurable H3 UNET, Video VAE, Audio VAE and CLIP models
• Live NVIDIA GPU and VRAM monitoring
• Automatic LM Studio model unloading after prompt generation
• Manual ComfyUI model unloading
• Rendered video preview and metadata
• Windows startup script
• Clean first-run configuration

Installation

Please follow the installation instructions in README.md.

Before using ComfyMax, first make sure the supplied example workflow in
workflow_example renders successfully directly inside ComfyUI.

Status

This is the first public release. The included workflows have been tested
with one, two and three reference images.

License

Released under the MIT License.