A LoRA that softens the harsh upscale Qwen Image 2.1 does by itself: the result is a little softer, still detailed, makes up less, and stays true to your picture.Qwen Image 2.1 can upscale a small or worn picture without this LoRA, but it overdoes it. Skin turns gritty, pores and freckles show up that were not there, faces look older, and the picture shifts a few pixels. With this LoRA the result stays on your picture.Blurry photos need a lower strength. If your picture is out of focus, shaky, or a frame from a video, set the LoRA to 0.5. At 1.0 it stays true to the blur and the result comes back soft. A small or compressed picture that is in focus is fine at 1.0.No trigger word. Start at strength 1.0, enlarge your picture to about 2 megapixels, give it as image 1, and ask:<code>Restore the photo: remove blur, noise and compression artifacts and recover sharp, natural detail, keeping the same composition, people and colors.</code>For scratched or faded prints:<code>Restore the old photo: remove scratches, dust, fading and grain and recover sharp, natural detail and color, keeping the same composition and people.</code>On an old print it takes out scratches, dust and grain and brings back most of the color.<h3 id="why-use-it">Why use it</h3>Measured on 37 small pictures it never trained on, each compared with the full-size picture it was shrunk from. Same seed and request, 8 steps with Viggle Turbo:<ul><li>The picture stays in place. Without the LoRA the result sits more than 2 px off. With it: about 0.3 px.</li><li>The face stays the same person. Face match 0.79 without, 0.88 with (on 29 of them; closer on every one).</li><li>The colors stay. Color and brightness shift 1.9 % without, 0.9 % with.</li><li>It makes up less. Plain Qwen draws more grain than the full-size picture has. The LoRA draws a little less than it has, so results are slightly soft.</li></ul>The pictures on this page show it: the small input, the result without the LoRA (amber) and with it (teal), and a section enlarged underneath.<h3 id="how-to-use-it">How to use it</h3>The workflow on this page has it set up: load your image and press Run. One pass at about 2 megapixels, with a Fast switch (8 steps with Viggle Turbo, or 25 steps without it). It is the second file of this version, and it is inside every still picture here: drop one on ComfyUI. It uses my <a target="_blank" rel="ugc" href="https://github.com/ausboss/ComfyUI-AusBoss">AusBoss nodes</a>.For a result over 2 megapixels, use my <a target="_blank" rel="ugc" href="https://civarchive.com/models/2988629/qwen-image-21-tiled-upscale">Qwen Image 2.1 Tiled Upscale</a> workflow. It redraws the picture in tiles with this LoRA.In your own Qwen Image 2.1 edit graph:<ol><li>A LoRA loader right after the model loader, strength 1.0 to start. Strength works as a dial: 0.5 for a blurry photo, 0.75 for a little more of Qwen's own texture, 1.0 to stay truest to your picture.</li><li>Resize your picture to about 2 MP, both sides divisible by 32.</li><li>Text Encode Qwen Image 2.1: the resized picture as <code>image_1</code>, <code>resolution</code> 0, the request as the prompt.</li><li>KSampler on the encoder's <code>latent</code> output: CFG 1, <code>euler</code> / <code>simple</code>, denoise 1. 8 steps with Viggle Turbo at 1.0, or 25 steps without it.</li><li>VAE Decode, then Split Image with Alpha (the Qwen 2.1 VAE decodes RGBA).</li></ol><h3 id="limits">Limits</h3><ul><li>A blurry photo comes back soft at 1.0. Lower the strength to 0.5. On 26 real low-resolution photos, the 6 blurriest needed that.</li><li>It is a redraw, not a recovery. Fine detail is made up to fit, and a face stays the same person, not the same pixels.</li><li>Plain Qwen 2.1 draws more texture. This LoRA trades some of that for staying true to the picture.</li><li>Small faces are redrawn. A face about 100 px tall in your picture can come back as a slightly different person.</li><li>Trained at 1 to 2 megapixels. Keep each pass at about 2 MP; for a bigger result, redraw the picture in tiles (my <a target="_blank" rel="ugc" href="https://civarchive.com/models/2988629/qwen-image-21-tiled-upscale">Qwen Image 2.1 Tiled Upscale</a> workflow does that).</li><li>Small lettering can come back misspelled.</li><li>One pass is enough. A second pass over the result at the same size came back crunchy.</li><li>Don't stack it with my Consistency LoRA.</li></ul><h3 id="training">Training</h3><ul><li>ai-toolkit (<code>qwen_image_2</code>) on the Comfy-Org INT8 convrot base. Rank 32, alpha 32, AdamW8bit, learning rate 1e-4, batch 1, 1,500 steps.</li><li>1,209 pairs: a clean picture, and the same picture damaged the way pictures get damaged (shrunk for the web, saved and re-saved as JPEG, blur and noise, phone smoothing, old prints, video stills).</li></ul>The full numbers, the step 1,250 file and the notes on how it was made are on the <a target="_blank" rel="ugc" href="https://huggingface.co/ausboss/Qwen-Image-2.1-Restore-LoRA">Hugging Face model card</a>.Qwen Image 2.1 is by Alibaba Qwen, under the Qwen Research License: non-commercial use only.Questions or bugs: the comments here or <a target="_blank" rel="ugc" href="https://github.com/ausboss/ComfyUI-AusBoss/issues">GitHub issues</a>. More from me on <a target="_blank" rel="ugc" href="https://x.com/Zanzibased">X @Zanzibased</a> and <a target="_blank" rel="ugc" href="https://github.com/ausboss">GitHub</a>.

Qwen Image 2.1 Restore LoRA

Step 1,500. Step 1,250, with a little more skin texture, is on <a target="_blank" href="https://huggingface.co/ausboss/Qwen-Image-2.1-Restore-LoRA" rel="ugc">Hugging Face</a>.

Files with SHA256: 39160f2021884492e590f8295bedb236dc26e13b7d9ee2bb824c03011f592351