BlackMixture/Mix-Studio

Mix Studio is a free, open-source app that gives you a clean, touch-friendly web interface for running AI image and video generation on your own computer. It's built on top of ComfyUI, a popular tool for AI media generation, but instead of making you build complex node graphs, Mix Studio gives you focused controls and assembles the underlying workflow for each job automatically.

What it does:

• Generates images (Create workspace) with models like Krea 2, Flux 2 Klein, and Qwen Image 2.1, including controls for resolution, seed, batching, and LoRA (style) adjustments.
• Regional prompting (Region workspace): draw boxes on a canvas, give each box its own prompt, style stack, and reference image, and Mix Studio merges them into one coherent image.
• Edits images (Edit workspace): whole-image editing, inpainting/outpainting, multi-reference editing (you can type @Image 1 to reference a specific uploaded image as a prompt token), and identity-preserving edits.
• Generates video (Video workspace): text-to-video, first/last-frame video, reference-to-video, lipsync, character animation, and motion transfer, using models like MiniMax H3, LTX 2.3/2.5, Wan 2.2, and SCAIL 2.
• Upscales images with a synchronized zoom/pan comparison view.
• Experimental 3D: turn a photo into an orbit video and then a Gaussian splat (a 3D point-cloud representation), viewable locally.
• Library: searchable storage for all your generations and uploads, with folders, groups, metadata, reusable settings, trash recovery, and ZIP export.

Key design principles:

• Local and private: prompts, inputs, generated media, profiles, and galleries stay on your computer. Nothing is sent to a cloud service.
• Reuses your existing ComfyUI: Mix Studio detects your ComfyUI installation and uses your models, custom nodes, queue, and recovery outputs.
• Works from anywhere on your network: you can use the same interface from a phone on the same Wi-Fi, or privately through Tailscale.
• Profiles: separate local workspaces with optional PIN access.

Hardware requirements:

• Windows with NVIDIA supports the full curated workflow set.
• Linux with NVIDIA works via an existing source-based ComfyUI environment.
• Apple Silicon macOS supports a subset (Metal-compatible) with official BF16 LTX 2.3 weights; some NVIDIA-specific routes (LTX 2.5, MiniMax H3) are unavailable there.
• AMD ROCm on Windows/Linux is experimental.
• The lowest guided route is a 4 GB offloaded edit workflow; 16 GB VRAM is the practical recommendation for images, 24 GB for larger video workflows.

Setup:

• Download a Windows or macOS installer from the project's download page, or clone the repository on Linux and run ./start-mixstudio.sh.
• Mix Studio opens in your browser, detects an existing ComfyUI installation, and guides you through the files required by the workflows you choose.
• Updates are handled in-app (side menu → Update app), which fast-forwards when the Git checkout is clean and queues are idle.

License: GNU General Public License v3.0. The GPL applies to Mix Studio's own source code; ComfyUI, model weights, and third-party components keep their own licenses. Mix Studio does not claim ownership of media users create.

Attribution: ComfyUI executes the API graphs; model creators include Black Forest Labs, MiniMax, Lightricks, Krea AI, and the Wan team; community projects like SCAIL 2, 10Eros, SeedVR2, and Ultimate SD Upscale provide specialized components. Development was done on a Dell Pro Max T2 Tower with an NVIDIA RTX PRO 6000 Blackwell GPU (96 GB VRAM).