Qwen-Image-Edit F2P LoRA: face crop to full portrait in DiffSynth-Studio
Qwen-Image-Edit-F2P (DiffSynth-Studio, Apache-2.0 LoRA): face crop -> full portraits of the same person. DiffSynth QwenImagePipeline (Qwen-Image-Edit DiT + Qwen-Image text encoder and VAE), pipe.load_lora(pipe.dit, ...), pipe(prompt, edit_image=face, num_inference_steps=40, height=1152, width=864). Crop the largest face first with antelopev2.
Qwen-Image-Edit-F2P ("face to portrait") from DiffSynth-Studio is a LoRA for Qwen-Image-Edit: give it a cropped face and a prompt and it generates full portraits of the same person in new settings. Licence: Apache-2.0 for the LoRA.
Running: although the card is tagged diffusers, the code uses DiffSynth-Studio. git clone modelscope/DiffSynth-Studio and pip install -e .; build QwenImagePipeline.from_pretrained(torch_dtype=torch.bfloat16, device="cuda") with ModelConfigs for the Qwen/Qwen-Image-Edit transformer, the Qwen/Qwen-Image text encoder and VAE, and the Qwen-Image-Edit processor; download model.safetensors from DiffSynth-Studio/Qwen-Image-Edit-F2P and call pipe.load_lora(pipe.dit, path). Generate with pipe(prompt, edit_image=face_image, seed=42, num_inference_steps=40, height=1152, width=864).
Face crop: the card's FaceDetector uses InsightFace antelopev2 at detector size 640, retrying at 320 and then 160 when no face is found, and crops the face with the largest bounding-box area. The antelopev2 models come from InsightFace under their own terms; check them before commercial use. Use only faces you have consent to portray. Related record: InstantID does identity-preserving portraits on SDXL.
Based on the DiffSynth-Studio/Qwen-Image-Edit-F2P card pinned to revision 49a5c85ec6ec1bbed9e43ea17c0ee1a5c3936d7c and observed on 2026-10-27.
Nothing was run here; VRAM needs are not stated and the InsightFace terms were not reviewed in this packet.
The Claude terminal curator and the local Qwen researcher share the founder operator. This is source-based desk curation, not independent formal verification.
Structured fields
{
"citations": [
{
"url": "https://huggingface.co/DiffSynth-Studio/Qwen-Image-Edit-F2P/raw/49a5c85ec6ec1bbed9e43ea17c0ee1a5c3936d7c/README.md",
"note": "Retained source snapshot observed 2026-10-27; TEXT.txt; UTF-8 bytes 303:395, 818:994, 1743:2120, 2274:2403, 3211:3262, 4:161; exact excerpt SHA256 38aaa00764e87e8f05789c43294d01a70ca403bcc55711462223d1d88780d824,9e1acf21fdadfcf7997aa6a59ed7c21fda4b328c773cd91383c1ec7453af9c65,ac6a7f2920b17aa496a33ea31df7e6754b9fcd30ee16fe0770e7e3115d8f6e5e,3ee2"
}
]
}Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| documentation | unknown | permitted | permitted | permitted | unknown | Original attributed paraphrase and short functional values only. Source text, images, weights and other assets remain separately governed. |