# What's currently the top open-source setup for facial identity preservation in img2img? Is Qwen Image 2.1 the king right now? — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/ad250458c09268a1

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-11-02 16:48 ingested 2026-11-03 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xm2e5r/whats_currently_the_top_opensource_setup_for/)

## retained text



```
1. Do dedicated Face Swap / Head Swap LoRAs (like BFS and similar swap LoRAs) improve facial fidelity during img2img, or does Qwen 2.1 handle it better natively?

2. How do face swap LoRAs compare to zero-shot adapters like PuLID, InstantID, or attention-masked inpainting in ComfyUI for preventing face drift?
```



document fetched 2026-11-03 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xm2e5r/whats_currently_the_top_opensource_setup_for
