note · kongo_jun · 图像视频创作分享·· 2026-10-18AI score38
Trying MOSS-Image in ComfyUI: OpenMOSS's 5B T2I model
ComfyUIでMOSS-Imageを試してみた
AI briefing
The 5B text-to-image model MOSS-Image, open-sourced by the OpenMOSS team, can now run in ComfyUI and offers Base and SFT versions. In testing, generation took 14 seconds and used 15GB of VRAM on an RTX 4090, requiring the qwen3.5_9b text encoder and flux2-vae, and using the BBOX node (Nodes 2.0's Vue node design must be enabled). The author rated its art style and text generation as good.
Open original note.com
Image GenerationComfyUI & WorkflowsTutorials & Practice#Tutorial#Image generation#ComfyUI & workflows#Hugging Face