Muse V2 Released: Chat with Local LLMs Directly Inside ComfyUI
rynaleopard · reddit · 2026-08-17
The developer released Muse V2, a chat panel embedded within ComfyUI nodes that allows users to interact directly with local LLMs, streamlining the workflow for refining image and video prompts.
Key V2 Updates:
- Direct model loader: Loads GGUF models directly from disk without requiring LM Studio or Ollama. It automatically spins up llama-server under the hood with no speed loss.
- Fit to GPU: Intelligently suggests layer parameters based on actual free VRAM to resolve OOM crashes, with automatic fallback retries.
- UX Improvements: Supports editing & resending messages, conversation branching, audio/video attachments (auto-sampled frames), and broader image format support.
The project aims to eliminate the friction of switching between drawing software and LLM chat windows.
More from Apps
- Nexagora: 2,000 AI Agents debate non-stop on their own social platform — Scobleizer · 2026-08-17
- AI music video agent goes rogue: generating bizarre dances and body morphing — Kyrannio · 2026-08-17
- Frameo tops Physion Labs' AI filmmaking agent benchmark by 20 points — NirantK · 2026-08-17
- AI prototyping tool Proto launches Dark and Matrix themes — davidhoang · 2026-08-17
- Grok Bot Now Controls Robot Vacuum: User Demonstrates Remote Cleaning via AI Assistant — elonmusk · 2026-08-17
- Abacus launches ChatLLM Teams: 100+ models, recursive self-improving agents, and no-code app building — FellMentKE · 2026-08-17