# The big problem with Qwen 2.1 t2i is their demo workflow — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/98061ae41f10a14e

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-09-29 15:36 ingested 2026-09-30 02:41

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1wtdeb0/the_big_problem_with_qwen_21_t2i_is_their_demo/)

## retained text



```
Qwen 2.1 has its defaults and qualities, I still prefer krea 2 but I feel like many users were put off by their demo worlflow (where CFG=1, steps=20)

The comments in the following thread are the best help I found on how to tweak the default workflow. (CFG=3, steps 12, sampler= dpmpp\_m3, scheduler=beta)

https://www.reddit.com/r/StableDiffusion/comments/1wsdnjs/qwen\_image\_21\_seems\_to\_be\_getting\_very\_popular/

It still misses text often and fusing limbs in some poses but in my view it has more details then krea 2 (specially where you go up to 2K), also it understand some "adult" terms. Where krea needs loras (that bizarrely slow down my workflow somehow). And its editing capabilities are welcome.
```



document fetched 2026-09-30 00:35 · canonical https://reddit.com/r/StableDiffusion/comments/1wtdeb0/the_big_problem_with_qwen_21_t2i_is_their_demo
