ComfyUI node feeds reference images through MOSS-Image's vision tower as text conditioning
ThetaCursed · reddit · 2026-10-21
Reddit user ThetaCursed released a custom ComfyUI node and workflow for MOSS-Image that encodes reference images through the Qwen3.5-9B vision tower and injects the embeddings into the text conditioning — so the image guides generation as part of the prompt rather than as a VAE latent. The node exposes imageweight and visionmegapixels controls, and the demo workflow pairs it with the MOSS-Image-SFT-8-steps-dmd LoRA for 8-step sampling, reproducing the reference composition closely. Code and workflow are open-sourced on GitHub.
Related event: MOSS-Image Found to Accept Image Input, Paving Way for Image Editing(3 posts)→
More from coding & agent
- Blinded agents compare formal verification vs code review: empirical results are in, and they're striking — rickasaurus · 2026-10-24
- GitHub ships HydraFusion Max, orchestrating Claude Opus 5.5 and GPT-6 models into one response — unixterminal · 2026-10-24
- kimo: an open-source no-GUI video editor that coding agents drive over MCP — LieTop7965 · 2026-10-24
- tinyfool builds a personal shell app: fold remote output, edit remote files locally with auto-diff — tinyfool · 2026-10-24
- Dev shares how they built a Claude client for the iPhone Duo dual-screen setup — rudrank · 2026-10-24
- Hanging Context: keyless MCP server for citable news lands in Claude's directory — bshivarthy · 2026-10-24