Tired of the MOSS Image SFT's aggravated GPT look? You are at the right place This is a by-product of <s>coping with bumpy </s>moss image reference training attempts Re-sample your MOSS Image SFT renders to reduce the glossy, GPT-style look: less sheen and blush, flatter shading, more defined line work. The character and composition stay mostly the same.This is a second pass, not a normal style LoRA. Generate with MOSS SFT first, then run the image through this LoRA with the instruction below. Loading it during normal text-to-image changes almost nothing.<h2 id="how-to-use">How to use</h2>No custom nodes are needed. Use a current ComfyUI that includes the Text Encode MOSS-Image Edit node.<ol><li>Put this LoRA in <code>models/loras/</code> and name the file <code>ungpt_v1.safetensors</code>.</li><li>Get the MOSS SFT model, text encoder and VAE from <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MOSS-Image">Comfy-Org/MOSS-Image</a>.</li><li>Download a workflow from <a target="_blank" rel="ugc" href="https://github.com/bag3labo/moss-image-ungpt/tree/main/workflows">GitHub</a> and drag it into ComfyUI:<ul><li>moss_sft_t2i_then_ungpt.json: from prompt to final image, both passes in one queue.</li><li>ungpt_resample_image.json: re-sample an image you already have.</li></ul></li></ol><h2 id="settings-to-keep">Settings to keep</h2>These are the settings the results below were measured with. The workflows already use them.<ul><li>Instruction: Redraw this image as hand-drawn anime art without the AI-generated look. Keep the character, pose, outfit and framing.</li><li>Negative: empty prompt, with the same image attached</li><li>Edit format: ai-toolkit (kv_cache off)</li><li>LoRA strength: 1.0</li><li>Sampling: 50 steps, CFG 3, euler, sgm_uniform, denoise 1.0</li><li>Guidance: Skip Layer Guidance DiT Simple, single_layers 22, double_layers empty</li><li>Size: about 1 MP, sides in multiples of 16, output the same size as the input</li></ul>The re-sample starts from empty noise and is guided by the input image, so denoise stays at 1.0. It is not a low-denoise img2img pass.On an RTX PRO 6000, text-to-image took about 20 s and the re-sample about 46 s, with about 28 GB of VRAM in use. Smaller cards and the quantized model files were not tested.<h2 id="results">Results</h2>Tested on 24 text-to-image prompts with one seed each. The score comes from a classifier that separates gpt-image-2 images from hand-drawn anime art (lower is less GPT-like, under 0.5 reads as hand-drawn).<ul><li>MOSS SFT text-to-image: median 0.927, 3 of 24 under 0.5</li><li>With the LoRA loaded during text-to-image: median 0.909, 5 of 24 under 0.5</li><li>Re-sampled without the LoRA: median 0.952, 4 of 24 under 0.5</li><li>Re-sampled with the LoRA: median 0.452, 13 of 24 under 0.5</li></ul>The re-sample lowered the score on all 24 images. The same character was kept in 24 of 24 and the pose in 22 of 23 where a pose could be detected. The classifier never saw MOSS images in training, so treat its scores as a guide and judge the images yourself.<h2 id="limits">Limits</h2><ul><li>The effect varies. On 4 of the 24 images the score moved by 0.04 or less. The last sample image is the smallest change in the test.</li><li>Details can change. Expect small changes to hair, faces and expressions, and check any text in your output.</li><li>It adds a look, not an artist. It moves images toward a generic hand-drawn style and does not recover any specific artist's style.</li><li>gpt-image-2 input is less reliable. The median score fell from 0.890 to 0.013 on 36 test images, but the pose was kept in only 83%. Colour casts can get stronger, and framing and faces can drift more than on MOSS images.</li><li>Not tested: the MOSS Image Base model, more than one seed per prompt, other LoRA strengths or samplers, and input from other models.</li></ul>The classifier score is not an AI detector, and this LoRA has not been tested against one.Full details, workflows and license (Apache-2.0): <a target="_blank" rel="ugc" href="https://github.com/bag3labo/moss-image-ungpt">github.com/bag3labo/moss-image-ungpt</a>

Files with SHA256: 053a2df9828ed32df0287da4429e55811ba92694c7cb5a35eb21f9b1a8d2aa26