Extend a picture on any side, and Qwen Image 2.1 paints the new area. Qwen3-VL writes the prompt, and my outpaint LoRA keeps your picture in place. Your image goes back on top of the result.I put FLUX 3 Klein and Qwen Image 2.1 up against Nano Banana and ChatGPT on the same edits: <a target="_blank" rel="ugc" href="https://youtu.be/TJCG2kBRcDE">the video on YouTube</a>.<h2 id="how-to-use-it">How to use it</h2><ol><li>Load your image.</li><li>Pick a ratio, or drag the orange handles out. The small button next to the ratios flips portrait and landscape.</li><li>Optional: finish the sentence in What goes there to say what belongs in the new area.</li><li>Press Run.</li></ol><h2 id="what-you-need">What you need</h2><ul><li>AusBoss nodes 2.5.0 or newer. In ComfyUI-Manager, search "AusBoss". Already have them? Update first.</li><li>A ComfyUI build with Text Encode Qwen Image 2.1 (mid-September 2026 or newer).</li><li>The Qwen Image 2.1 models and <a target="_blank" rel="ugc" href="https://civarchive.com/models/2967375">my outpaint LoRA</a>:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/diffusion_models/qwen_image_2.1_int8_convrot.safetensors">qwen_image_2.1_int8_convrot.safetensors</a> → <code>models/diffusion_models</code> · 7.26 GB</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors">qwen3vl_8b_int8_convrot.safetensors</a> → <code>models/text_encoders</code> · 9.35 GB</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors">qwen_image_2.1_vae_bf16.safetensors</a> → <code>models/vae</code> · 676 MB</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/ausboss/Qwen-Image-2.1-Outpaint-LoRA/resolve/main/qwen-image-2.1-outpaint-v2.safetensors">qwen-image-2.1-outpaint-v2.safetensors</a> → <code>models/loras</code> · 159 MB</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/ausboss/Qwen-Image-2.1-Outpaint-LoRA/resolve/main/qwen-image-2.1-outpaint.safetensors">qwen-image-2.1-outpaint.safetensors</a> → <code>models/loras</code> · 159 MB (v1, for big zoom-outs)</li></ul></li></ul><h2 id="tips">Tips</h2><ul><li>Keep the LoRA on. Without it Qwen often shifts your picture. If the file is missing, the run stops at a red LoRA Loader; switch its row off to run without it.</li><li>Say what's missing. The box starts with <code>Also include in the prompt:</code>. Add to it, for example <code>she wears blue jeans.</code> If you delete those words and type your own, the workflow puts them back.</li><li>Pull out at most about your picture's own size on one side per run. More than that can paint a second scene. For more, run again on the result.</li><li>Crop out edges that confuse it. Noise, a frame line or a busy strip along the edge of your picture can make the new area come out strange. Drag the blue handles on Image Crop + Rotate + Pad inward to cut those edges off first.</li><li>Dark corners get extended. A vignette in your photo carries on into the new area.</li><li>Transparent PNG? A person or a scene cut out works. A product cut out with nothing behind it can come back on plain gray, so put it on a background first.</li><li>About 8 seconds a run on an RTX 5090.</li></ul>Tilted picture, or something to crop away? Try <a target="_blank" rel="ugc" href="https://civarchive.com/models/2966883">Qwen Image 2.1 Rotate + Outpaint</a>.<h2 id="credits">Credits</h2>Qwen Image 2.1 is © Alibaba Qwen, under the Qwen Research License: non-commercial use only, and commercial use needs a license from Qwen. The outpaint LoRA is mine and follows that license. I trained it on my own pictures plus CommonCatalog CC-BY, PD12M and anime-with-caption-cc0.v1.2 (2026-09-30): the padding follows your picture, the seam blends in, and your own words always reach the prompt. v1.1 (2026-09-26): a missing LoRA stops the run, LoRA v2 by default. v1.0 (2026-09-25): first release.Questions or bugs: <a target="_blank" rel="ugc" href="https://github.com/ausboss/ComfyUI-AusBoss/issues">GitHub issues</a> or the comments here. Videos on <a target="_blank" rel="ugc" href="https://www.youtube.com/@ausbossdev">YouTube @ausbossdev</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 Outpaint | Auto Prompt

<ul><li>The padding now follows your picture. Pick a ratio or drag the orange handles, and the padding is worked out for whatever image you load. Before, it was saved in pixels for the sample, so another picture came out wrong.</li><li>Image Crop + Rotate + Pad replaces Load Image + Pad, which gives you the ratio buttons.</li><li>The top level is smaller: 13 nodes instead of 23, with the prompt and the painting in two boxes.</li><li>What goes there now starts with "Also include in the prompt:", so you only finish the sentence. Before, it held a line about the sample's skirt.</li><li>Your own words always reach the prompt. If you delete "Also include in the prompt:" and type your own line, the workflow puts those words back, so a line like "a sunny beach behind her" now shows up.</li><li>The seam is set to blend in. On a turned or dark picture the old seam could leave a line along your picture's edge.</li><li>Transparent PNGs: the prompt writer now sees the see-through parts in the fill colour instead of black, so it stops describing a black frame or corners.</li><li>Needs AusBoss nodes 2.5.0 or newer. The models are the same.</li></ul>

Files with SHA256: f566545cea651dbbf4f292cd83ece8804353f962a5bcc791c91f7ecd642567da