A high-resolution MOSS-Image text-to-image workflow using the MAX ULTRA LoRA. The active graph starts from a 2560 x 1440 latent canvas and saves a generated image.
Main features:
- MOSS-Image Base INT8 ConvRot with moss_hires_lora_test_v0.4.safetensors at strength 1, plus Qwen3.5 9B BF16 text encoder and FLUX.2 VAE.
- One output-reachable Euler sampler at 40 steps, CFG 3, simple scheduler, fixed seed 9527.
- Text prompt and negative prompt feed the generation path; no input image or tiled upscaling stage is connected.
Suggested workflow:
1. Install the model, LoRA, text encoder, and VAE named in the JSON, then load it in ComfyUI.
2. Edit the prompts and generate at the supplied 2560 x 1440 canvas.
3. Inspect composition and fine detail at full size; reduce dimensions if your hardware cannot complete the run.
This is direct text-to-image generation, not enlargement of an uploaded picture. The paired cover is promotional artwork, not a measured output of this JSON. Older notes inside the workflow mention different settings; the connected sampler and latent canvas above are the current configuration.
RunningHub AI workflow: https://www.runninghub.ai/zh-cn/post/2114399092683112450?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