# Qedit 2511 seems better then 2.1 — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/14c383ca4fc195f0

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-11-09 11:57 ingested 2026-11-10 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xrpshw/qedit_2511_seems_better_then_21/)

## retained text



```
I'm just a beginner in comfyui and stuff with very limited hardware.

I have rtx 3050 4gb vram and 8gb ddr4 single channel

Running comfyui with kitchen attention

I ran qedit 2511 fp8 with 4 step lighting Lora with default scheduler or sampler and it took around 110 seconds

And qwen 2.1 turbo int8(runs on 8 steps) without the prompt enhancement took 90 seconds 

Now I didn't use any prompt enhancement with qedit but it still worked fine the biggest issue was the facial drift 

In qwen 2.1, the facial drift has reduced drastically (as much as I have seen) but the prompt adherence is really bad. 

I used the original model they provided for enhancement and it was shit + it was restricted. Since I have limited hardware I tried qwen 3.5 2b abliterated. It works sometimes but doesn't a lot of time 

Seeing how qedit literally didn't need a enhancer and still worked fine and 2.1 isn't working well

I feel like it's mostly cuz of the enhancer but even when I wrote a good prompt without using enhancer it would still give shit results 

Should I just use qedit? The facial drift is amazing bad tho (because of turbo Lora ofcourse but my hardware 🥀) 
```



document fetched 2026-11-10 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xrpshw/qedit_2511_seems_better_then_21
