This is a private version trained on larger training parameters. Compared with the beta version, it delivers better performance in detail rendering, generalization and high-resolution output. This version is available only to my supporters. Resale is strictly prohibited. Please let me know if you find anyone selling it.

Preface

This is probably the most difficult version among all the HighQuality models I've trained. I've gone through no fewer than ten iterations in total, adjusting training parameters and datasets along the way, even scrapping the original dataset entirely and building a new one from scratch, but none of them met my expectations until this version came along. That's why this release is a bit later than planned.By the way, let's talk about the features of QI-21: it performs quite well in terms of consistency, especially the consistency of human faces. Tools like the previous Krea2Edit could barely achieve this, and I had to use a workflow to forcibly suppress its urge to modify images, but there's no such worry at all with QI-21 — that's its advantage. Now for its drawbacks, there are two points: first, the images generated by QI-21 always give an "unconverged" feeling, as if they haven't been fully generated yet; second, it produces relatively noticeable noise. To address this, I've specifically created a dedicated workflow for optimization, which you can access here.

Alright, now let's talk about today's star product: ✨HighQuality for QI-21✨

Similarly, the private version is trained based on the beta version with higher training parameters, so it performs better on images with higher resolution and delivers richer details.

Parameters:

Sampler : euler/res_multistep

Scheduler : beta/sgm_uniform

Steps : 8 ( turbo )

CFG: 1

Trigger words : no

Exclusive Workflow :

QI-21 Single Image Editing Workflow

Solve the greasy image issue of QI-21

The scaling workflow of QI-21 supports a maximum of 32K

Example:

This model has been specially optimized for human figures, especially for skin and hair strands.

Multi-character Optimization

Optimization of Special Lighting and Shadows

Product Optimization

Optimization of Film and Television Screenshots

✨Member Event✨:

This LoRA is exclusive to LV3/LV4 users. Users who are not LV3/LV4 need not be disappointed. I will randomly select members every week to give away the private version of this LoRA. Whether you are an LV1 or LV2 user, leave a comment "✨HighQuality for QI-21✨" under this post now, and you will stand a chance to receive this gift. Hopefully you will be the next lucky winner.

OK! That's all! I hope you like it!

If you enjoy my sharing, you are welcome to join my private club. If you are already a member, you can get the project files for this sharing directly below. 👇👇👇

🫡May the AI-power be with you, see you soon!🫡

Disclaimer: My model or workflow is solely for learning and communication purposes and shall not be used for any commercial activities. All responsibilities for the generated content, including its legality, ethical implications, or potential misuse, are borne entirely by the user. By using this model or workflow, you agree to this principle. Please use it responsibly within the scope of local laws.

6OWCdeAQGa64JbH9J0tzVC4r3hO3tcJCwpr4-TP7ERA

Click here