LoRA-Edit: Controllable First-Frame-Guided Video Editing via Mask-Aware LoRA Fine-Tuning

https://redd.it/1la6nta
@rStableDiffusion

Anyone knows how to create this art style?
https://redd.it/1la5t4c
@rStableDiffusion

Looking for alternatives for GPT-image-1

I’m looking for image generation models that can handle rendering a good amount of text in an image — ideally a full paragraph with clean layout and readability. I’ve tested several models on Replicate, including imagen-4-ultra and flux kontext-max, which came close. But so far, only GPT-Image-1 (via ChatGPT) has consistently done it well.

Are there any open-source or fine-tuned models that specialize in generating text-rich images like this? Would appreciate any recommendations!

Thanks for the help!

https://redd.it/1la8fex
@rStableDiffusion

Any clue what causes this fried neon image?
https://redd.it/1la9ogx
@rStableDiffusion

Demo of WAN Fun-Control and IC-light (with HDR)
https://www.youtube.com/watch?v=bwVEuWmfsFA

https://redd.it/1la9jdh
@rStableDiffusion

I’ve made a Frequency Separation Extension for WebUI

https://redd.it/1lad68p
@rStableDiffusion

NexFace: High Quality Face Swap to Image and Video

I've been having some issues with some of popular faceswap extensions on comfy and A1111 so I created NexFace, a Python-based desktop app that generates high quality face swapped images and videos. NexFace is an extension of Face2Face and is based upon insight face. I have added image enhancements in pre and post processing and some facial upscaling. This model is unrestricted and I have had some reluctance to post this as I have seen a number of faceswap repos deleted and accounts banned but ultimately I beleive that it's up to each individual to act in accordance with the law and their own ethics.


Local Processing: Everything runs on your machine - no cloud uploads, no privacy concerns
High-Quality Results: Uses Insightface's face detection + custom preprocessing pipeline
Batch Processing: Swap faces across hundreds of images/videos in one go
Video Support: Full video processing with audio preservation
Memory Efficient: Automatic GPU cleanup and garbage collection
Technical Stack
Python 3.7+
Face2Face library
OpenCV + PyTorch
Gradio for the UI
FFmpeg for video processing
Requirements
5GB RAM minimum
GPU with 8GB+ VRAM recommended (but works on CPU)
FFmpeg for video support

I'd love some feedback and feature requests. Let me know if you have any questions about the implementation.

https://github.com/ExoFi-Labs/Nexface/

[Image Sample 1](https://i.imgur.com/w1pmVY2.png)

Image Sample 2

https://redd.it/1ladt87
@rStableDiffusion

How to Train Your Own LoRA in ComfyUI | Full Tutorial for Consistent Character (Low VRAM)
https://youtu.be/4idwy6NnJO8

https://redd.it/1lag4ey
@rStableDiffusion

A new way to play Phantom. I call it the video version of FLUX.1 Kontext.

https://redd.it/1laep3c
@rStableDiffusion

ByteDance just released a video model based off of SD 3.5 and Wan's vae.

https://redd.it/1lah1pv
@rStableDiffusion

Deeplive – any better models than inswapper128?

is there really no better model to use for deeplive and similar stuff than inswapper\128? its over 2 years old at this point, and surely theres something more recent and open source out there.

i know inswapper 256 and 512 exist, but theyre being gatekept by the dev, either being sold privately for an insane price, or being licensed out to other paid software.

128 feels so outdated looking at where we are with stuff :(

https://redd.it/1laek9v
@rStableDiffusion

Open Source V2V Surpasses Commercial Generation

A couple weeks ago I made a comment that the Vace Wan2.1 was suffering from a lot of quality degradation, but it was to be expected as the commercials also have bad controlnet/Vace-like applications.

This week I've been testing WanFusionX and its shocking how good it is, I'm getting better results with it than I can get on KLING, Runway or Vidu.

Just a heads up that you should try it out, the results are very good. The model is a merge of all of the best of Wan developments (causvid, moviegen,etc):

https://huggingface.co/vrgamedevgirl84/Wan14BT2VFusioniX


Btw sort of against rule 1, but if you upscale the output with Starlight Mini locally the results are commercial grade. (better for v2v)

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

PartCrafter - Have you guys seen this yet?
https://redd.it/1lam22x
@rStableDiffusion

Jib Mix Realistic XL V17 - Showcase

https://redd.it/1lanmz7
@rStableDiffusion