<div data-youtube-video="https://youtu.be/bvntROsEru0"><iframe src="https://www.youtube.com/embed/bvntROsEru0" width="853" height="480" allowfullscreen></iframe></div> 
 
Compare three connected MOSS-Image SFT generation paths in one ComfyUI graph. The branches share prompt, seed 4242, and a 1024 x 1024 latent canvas, then save separate 50-step, 8-step, and 4-step results. 
 
Main features: 
- MOSS-Image SFT INT8 ConvRot, Qwen3.5 9B BF16 text encoder, and FLUX.2 VAE across three output-reachable branches. 
- Standard branch: 50 Euler steps, CFG 3, simple scheduler, with active Skip Layer Guidance 22. 
- DMD2 branch: 8 Euler steps, CFG 1, simple scheduler, with 8-steps-dmd2-lora.safetensors at strength 1. 
- Turbo branch: 4 Euler steps, CFG 1, simple scheduler, with moss_image_sft_4step_v2.0.safetensors at strength 1. 
 
Suggested workflow: 
1. Install the SFT model, text encoder, VAE, and both acceleration LoRAs, then load the JSON. 
2. Run the shared prompt to produce all three saved images. 
3. Compare subject fidelity, detail, style, and actual run times; repeat with other prompts. 
 
This is a comparison of three recipes with different LoRAs and CFG values, not a single-variable step-count test. The paired cover is promotional artwork, not a measured output of this JSON. 
 
RunningHub AI workflow: https://www.runninghub.ai/zh-cn/post/2114400597616455681?inviteCode=rh-v1111 
RunningHub signup: https://www.runninghub.ai/?inviteCode=rh-v1111 
Bilibili video: https://www.bilibili.com/video/BV1UYnU6qEGK/ 
Related model files: https://pan.quark.cn/s/9ec095b95838

MOSS-Image 50/8/4-Step Comparison Lab

<div data-youtube-video="https://youtu.be/bvntROsEru0"></div> 
 
Compare three connected MOSS-Image SFT generation paths in one ComfyUI graph. The branches share prompt, seed 4242, and a 1024 x 1024 latent canvas, then save separate 50-step, 8-step, and 4-step results. 
 
Main features: 
- MOSS-Image SFT INT8 ConvRot, Qwen3.5 9B BF16 text encoder, and FLUX.2 VAE across three output-reachable branches. 
- Standard branch: 50 Euler steps, CFG 3, simple scheduler, with active Skip Layer Guidance 22. 
- DMD2 branch: 8 Euler steps, CFG 1, simple scheduler, with 8-steps-dmd2-lora.safetensors at strength 1. 
- Turbo branch: 4 Euler steps, CFG 1, simple scheduler, with moss_image_sft_4step_v2.0.safetensors at strength 1. 
 
Suggested workflow: 
1. Install the SFT model, text encoder, VAE, and both acceleration LoRAs, then load the JSON. 
2. Run the shared prompt to produce all three saved images. 
3. Compare subject fidelity, detail, style, and actual run times; repeat with other prompts. 
 
This is a comparison of three recipes with different LoRAs and CFG values, not a single-variable step-count test. The paired cover is promotional artwork, not a measured output of this JSON. 
 
RunningHub AI workflow: https://www.runninghub.ai/zh-cn/post/2114400597616455681?inviteCode=rh-v1111 
RunningHub signup: https://www.runninghub.ai/?inviteCode=rh-v1111 
Bilibili video: https://www.bilibili.com/video/BV1UYnU6qEGK/ 
Related model files: https://pan.quark.cn/s/9ec095b95838

Files with SHA256: 6aeb7b03109097144d27f40c100d2a608269e1cb439e76408fcd56a738f639c8