Loop Forge2048 pixels in one pass: the cafe sign
From MOSS-Image add-ons, tested locally
- Model
- MOSS-Image Base (moss_image_base_int8_convrot.safetensors)
- Output
- 2048×2048
- Sampling
- 25 steps, CFG 2.5, er_sde / simple
- Seed
- 101
- 2048 plain
- 235.5 s, peak VRAM 10.5 GiB
- Hi-res LoHa
- 238.2 s, peak VRAM 10.6 GiB
- Scale ROPE
- 239.7 s, peak VRAM 10.5 GiB
- Ran on
- RTX 3060, 12GB (local)
Both run on a 12 GB card, but it is a long wait for one image. They held together at that size, so if you need a really big picture, either one is worth the wait.
Photograph of a man with short black hair, a trimmed beard and a white linen shirt with rolled sleeves, standing in front of a small corner cafe at golden hour. Above the door a hand-painted wooden sign reads "THE SLOW KETTLE" and a chalkboard beside him reads "Fresh bread at 7. Open late on Fridays." Warm light, shallow depth of field, 35mm photo.
nsfw, worst quality, bad anatomy, watermark, jpeg artifact
Notes
The official Base template at 4 megapixels (2048×2048). The LoHa is comfyanonymous's hi-res test LoRA v0.5 at 1.0, for Base. Scale ROPE is a native ComfyUI node, set to 0.5 × 0.5 (NTK scaling from 1024 to 2048). 2048 plain has neither.
comfyanonymous/MOSS_image_MAX_ULTRA_hires_lora_test on Hugging Face (v0.5)
Comfy-Org workflow templates (the MOSS-Image templates ship with ComfyUI)
Run it yourself
This test loads no input files. Put the models below in place and drag the workflow onto ComfyUI.
| File | Put it in | Size |
|---|---|---|
| moss_image_base_int8_convrot.safetensors | models/diffusion_models | 5.4 GB |
| qwen3.5_9b_w4a8_gptq.safetensors | models/text_encoders | 7.3 GB |
| flux2-vae_small.safetensors | models/vae | 249.5 MB |
| moss_hires_lora_test_v0.5.safetensors | models/loras | 182.2 MB |
No custom nodes. Every node is part of ComfyUI 0.43.1 (frontend 1.57.6, templates 0.11.100).