A ready-to-edit MOSS-Image SFT text-to-image workflow for learning the standard 50-step generation path in ComfyUI. The connected graph goes from the SFT model and text encoding through one sampler, VAE decoding, and SaveImage.
Main features:
- MOSS-Image SFT INT8 ConvRot diffusion model, Qwen3.5 9B BF16 text encoder, and FLUX.2 VAE.
- One output-reachable Euler sampler: 50 steps, CFG 3, simple scheduler, fixed seed 20261021.
- A sample misty mountain valley prompt and a separate negative prompt. No LoRA is connected.
Suggested workflow:
1. Download the workflow JSON and obtain the required model files from the resource link below.
2. Place the model, text encoder, and VAE in their corresponding ComfyUI model folders, then load the JSON.
3. Edit the positive prompt and seed, run the graph, and review the SaveImage output. Adjust steps or CFG only after establishing a baseline.
Required model filenames: moss_image_sft_int8_convrot.safetensors; qwen3.5_9b_bf16.safetensors; flux2-vae.safetensors. Actual speed, memory use, and output quality depend on your hardware and model files.
RunningHub AI workflow: https://www.runninghub.ai/zh-cn/post/2113906843746922497?inviteCode=rh-v1111
RunningHub signup: https://www.runninghub.ai/?inviteCode=rh-v1111
Bilibili video: https://www.bilibili.com/video/BV1ZHHP6FEVZ/
Related model files: https://pan.quark.cn/s/20e19307e8ef