Wan2.2 FLF used for VFX clothing changes - There's a very interesting fact in the post about the Tuxedo.

https://redd.it/1on0v6v
@rStableDiffusion

Qwen Image Edit Lens conversion Lora test

https://preview.redd.it/bvwqoofaqzyf1.jpg?width=3666&format=pjpg&auto=webp&s=5090a938dbee41e249840760d7cbc3a3edecf4fa

https://preview.redd.it/q7gsql7hqzyf1.jpg?width=1970&format=pjpg&auto=webp&s=c55a1fd1db5080258a567ca1572829e42e55a543

Today, I'd like to share a very interesting Lora model of Qwen Edit. It was shared by a great expert named Big Xiong. This Lora model allows us to control the camera to move up, down, left, and right, as well as rotate left and right. You can also look down or up. The camera can be changed to a wide-angle or close-up lens.


**models link**：https://huggingface.co/dx8152/Qwen-Edit-2509-Multiple-angles

**Workflow down**：https://civitai.com/models/2096307/qwen-edit2509-multi-angle-storyboard-direct-output

The picture above shows tests conducted on 10 different lenses respectively, with the corresponding prompt: Move the camera forward.

* Move the camera left.
* Move the camera right.
* Move the camera down.
* Rotate the camera 45 degrees to the left.
* Rotate the camera 45 degrees to the right.
* Turn the camera to a top-down view.
* Turn the camera to an upward angle.
* Turn the camera to a wide-angle lens.
* Turn the camera to a close-up.

https://redd.it/1on560y
@rStableDiffusion

How do you curate your mountains of generated media?

Until recently, I have just deleted any image or video I've generated that doesn't directly fit into a current project. Now though, I'm setting aside anything I deem "not slop" with the notion that maybe I can make use of it in the future. Suddenly I have hundreds of files and no good way to navigate them.

I could auto-caption these and slap together a simple database, but surely this is an already-solved problem. Google and LLMs show me many options for managing image and video libraries. Are there any that stand above the rest for this use case? I'd like something lightweight that can just ingest the media and the metadata and then allow me to search it meaningfully without much fuss.

How do others manage their "not slop" collection?



https://redd.it/1on7h64
@rStableDiffusion

Flux Gym updated (fluxgymbuckets)

I updated my fork of the flux gym

[https://github.com/FartyPants/fluxgym\bucket](https://github.com/FartyPants/fluxgymbucket)

I just realised with a bit of surprise that the original code would often skip some of the images. I had 100 images, but FLux Gym collected only 70. This isn't obvious, only if you look in the dataset directory.
It's because the way the collection code was written - very questionably.

So this new code is more robust and does what it suppose to do.

You only need the [app.py](https://github.com/FartyPants/fluxgymbucket/blob/main/app.py) that's where all the changes are (backup your original, and just drop the new in)

Also as previously, this version also fixes other things regarding buckets and resizing, it's described in readme.

https://redd.it/1on8bcw
@rStableDiffusion

Telegram's Cocoon - AI network (Important)

Pavel Durov (Telegram's founder) has announced a new project called Cocoon.

It's a decentralized AI network built on the TON blockchain.
The goal is to let people use AI tools without giving up their data privacy to big tech companies.

https://preview.redd.it/qyr4pgb7c1zf1.png?width=700&format=png&auto=webp&s=6893fa20ea19738ebe2c137d553099479ab833f0



https://redd.it/1onachu
@rStableDiffusion

Finetuned LoRA for Enhanced Skin Realism in Qwen-Image-Edit-2509

Today I'm sharing a Qwen Edit 2509 based lora I created for improving Skin details across variety of subjects style shots.

I wrote about the problem, solution and my process of training in more details here on LinkedIn if you're interested in a bit of a deeper dive and exploring Nano Banana's attempt at improving skin, or understanding the approach to the dataset etc.



If you just want to grab the resources itself, feel free to download:

here on HF: [https://huggingface.co/tlennon-ie/qwen-edit-skin](https://huggingface.co/tlennon-ie/qwen-edit-skin)
here on Civitai: https://civitai.com/models/2097058?modelVersionId=2372630

The HuggingFace repo also includes a ComfyUI workflow I used for the comparison images.

It also includes the AI-Toolkit configuration file which has the settings I used to train this.


Want some comparisons? See below for some examples of before/after using the LORA.

If you have any feedback, I'd love to hear it. Yeah it might not be a perfect result, and there are other lora's likely trying to do the same but I thought I'd at least share my approach along with the resulting files to help out where I can. If you have further ideas, let me know. If you have questions, I'll try to answer.



https://preview.redd.it/yhph5r2vp1zf1.png?width=1333&format=png&auto=webp&s=fa4d649848cbf0061b50587a07785eeb79f94341

https://preview.redd.it/0wh4bs2vp1zf1.png?width=1333&format=png&auto=webp&s=aaafd02c6fae76b2f075efadee416379f1930afc

https://preview.redd.it/quhqxt2vp1zf1.png?width=1333&format=png&auto=webp&s=20d10493d681fca7998f0088d569d183e0e2f8f5

https://preview.redd.it/02ecc7xzp1zf1.png?width=3936&format=png&auto=webp&s=e5fe8690f77a9cc18a30b227fa3bbe924f50b910

https://preview.redd.it/3z6497xzp1zf1.png?width=3936&format=png&auto=webp&s=d166ed5d96295a2d4de4923c86c58ca05ef7b350

https://preview.redd.it/43q7ufxzp1zf1.png?width=3840&format=png&auto=webp&s=0a7f1bf052817b2770412c869daf631c027019e5

https://preview.redd.it/o6ab48xzp1zf1.png?width=3936&format=png&auto=webp&s=ebc94d0cc70dd8290c6fac69f3314f3c36d1a131

https://preview.redd.it/1o31e7xzp1zf1.png?width=4480&format=png&auto=webp&s=b67791b34e61272b977bfb195e8d6ec75745ae30

https://preview.redd.it/sy8557xzp1zf1.png?width=3936&format=png&auto=webp&s=bd7d2554ca5295dd45f48b86a4146410180eedbc

https://preview.redd.it/ce3yn8xzp1zf1.png?width=3936&format=png&auto=webp&s=aa75b4a735be9283d246d294d2899b0d9b906ef8

https://preview.redd.it/ahnq89xzp1zf1.png?width=3936&format=png&auto=webp&s=e7291843ae3404ff6c521f43b29689e572f1f973

https://preview.redd.it/52xgi8xzp1zf1.png?width=4096&format=png&auto=webp&s=ee72c3672c2534fe03848539684313e8f8a4f4a9

https://preview.redd.it/cz9ev9xzp1zf1.png?width=3936&format=png&auto=webp&s=550867f814d691b60e935a38c2970a958c2425a1



https://redd.it/1onc7ok
@rStableDiffusion

Psychedelic Animation of myself

https://redd.it/1ond7sa
@rStableDiffusion

Alibaba has released an early preview of its new AI model, Qwen3-Max-Thinking.

Even as an early version still in training, it's already achieving 100% on challenging reasoning benchmarks like AIME 2025 and HMMT. You can try it now in Qwen Chat and via the Alibaba Cloud API.

https://preview.redd.it/1r4kjj7je2zf1.png?width=680&format=png&auto=webp&s=1d0567f47199dc5cfda5d0c381b0e20da37c3f4a

https://redd.it/1onfljd
@rStableDiffusion

New node for ComfyUI, SuperScaler. An all-in-one, multi-pass generative upscaling and post-processing node designed to simplify complex workflows and add a professional finish to your images.
https://redd.it/1onh1mx
@rStableDiffusion

Do you think that in the future, several years from now, it will be possible to do the same advanced things that are done in ComfyUI, but without nodes, with basic UIs, and for more novice users?
https://redd.it/1onepyq
@rStableDiffusion

[Open Weights] Morphic Wan 2.2 Frames to Video - Generate video based on up to 5 keyframes
https://github.com/morphicfilms/frames-to-video

https://redd.it/1onhq7m
@rStableDiffusion

Made a small Warhammer 40K cinematic trailer using ComfyUI and a bunch of models (Flux, Qwen, Veo, WAN 2.2)

https://redd.it/1ong4ar
@rStableDiffusion

FreeGen beta released. Now you can create SDXL images locally on your iPhone.

https://redd.it/1onnsb1
@rStableDiffusion