deepbeepmeep

WanGP 10.3: You are all invited to LTX-2 Mega Mix Party Tons of new LTX2 Features: LTX2.3 Distilled 1.1, Outpainting Ic Lora, VBVR Lora, HQ mode, Id Lora (LTX with voice cloning), ... But Also: Ace 1.5 Turbo XL & Magi Human. Checkout this Outpainted 4:3 Video from a 16:9 video

Introducing Texturizer A gradio app that lets you: 1. Take any rigged mesh 2. Add any custom image 3. To generate a textured rigged mesh Needs around 8GB VRAM NVIDIA GPU, low VRAM powered by @deepbeepmeep's MMGP library + Hunyuan3D. 1-Click install.

@cars4ever Sorry I am not good at prompts, I use WanGP contextual Prompt Enhancer !

@deepbeepmeep Very nice 👍 Is there also something to help with script / prompt writing u can recommend?

WanGP 10.3: You are all invited to LTX-2 Mega Mix Party Tons of new LTX2 Features: LTX2.3 Distilled 1.1, Outpainting Ic Lora, VBVR Lora, HQ mode, Id Lora (LTX with voice cloning), ... But Also: Ace 1.5 Turbo XL & Magi Human. Checkout this Outpainted 4:3 Video from a 16:9 video

@MedOumoumad Time machine included too since new release is WanGP 11.3 not 10.3 !

Left: original 16/9 with black bars, Right: 4:3 Outpainted using a WanGP "Process Full Video" Plugin can outpaint full multi hours videos with low RAM & VRAM ! Many thanx to @MedOumoumad for the lora.

@hohastudios yep, that's the whole beauty of it. it is why deepy needs to deeply integrated in WanGP

@deepbeepmeep Will it off load vram from the agent when video generation happens?

Meet Deepy your friendly WanGP v11 Agent. It works offline with as little of 8 GB of VRAM and won't divulge your secrets. It is 100% free (no need for a ChatGPT/Claude subscription). You can ask Deepy to perform for you tedious tasks such as: Generate a black frame, crop a video, extract a specific frame from a video, trim an audio, ... Deepy can also perform full workflows including multiple models (LTX-2.3, Wan, Qwen3 TTS, ...). For instance: 1) Generate an image of a robot disco dancing on top of a horse in a nightclub. 2) Now edit the image so the setting stays the same, but the robot has gotten off the horse and the horse is standing next to the robot. 3) Verify that the edited image matches the description; if it does not, generate another one. 4) Generate a transition between the two images. or Create a high quality image portrait that you think represents you best in your favorite setting. Then create an audio sample in which you will introduce the users to your capabilities. When done generate a video based on these two files. github.com/deepbeepmeep/W…

WanGP 10.98 brings you LTX 2.3 Bigger model, Better audio & Image 2 Video, Support Higher Res, but thanks to WanGP LTX 2.3 still have the same ridiculously low VRAM requirements (6GB+ VRAM). github.com/deepbeepmeep/W…

Qwen3 TTS produces great multi lingual voice cloning. But doesn’t support emotions on cloned voiced. Index TTS 2 can do it all but is restricted to English & Chinese. With a bit of hacking you can combine these two models if you need voice cloning & emotions with other languages.

@deepbeepmeep Audio generation space is moving insanely fast. Having all these models in one tool is great for comparing outputs side by side. Which TTS model gives you the best results?

WanGP 10.9 is Breaking the Sound Barrier. Qwen3 TTS, Kugel Audio, HeartMula, Index TTS 2 (new!) are accelerated up to 10 times faster compared to their vanilla apps ! Most WanGP TTS models support now 2 speaker dialogue with Voice Cloning and Automated Dialogue Generation. And as usal all the gen audio files can be fed back to the many WanGP Video Generator models such as LTX-2 or Infinitalk. With Index TTS 2 you can easily varry speaker emotions by inserting simple tags in the prompt. Check this 100% AI generated (Index TTS2 + LTX-2) Oscar worthy performance:

@FuturistAgent Thx. Just send the tips to your local charities.

@deepbeepmeep WanGP v10.70 by DeepBeepMeep/ACE-Step v1.0 3.5B Wow! Where do I send the tips!

Let's Get Ready to Rumble ! Ace Step 1.5 & Kugel Audio 0 are in WanGP 10.7 to challenge Suno 5 & ElevenLabs ! github.com/deepbeepmeep/W…

@xclearcast I am not aware that sage 3 works better with any version of python on windows (it does with linux though). Please share a link to your info.

@deepbeepmeep Please update the Python environment to 3.12+. I want to use SageAttention 3, but it's not compatible with the current Python 3.10 setup. Thanks!

WanGP 10.5: Music for your hearts Heart Mula: Suno quality song with lyrics on your local PC. You can generate up to 4 min of music. Qwen 3 TTS: you can either do Voice Cloning, Generate a Custom Voice based on a Prompt or use a Predefined Voice Specialized Prompt Enhancers: if you enter the prompt in Heart Mula "a song about AI generation", WanGP Prompt Enhancer will generate the corresponding masterpiece for you. Likewise you can enhance "A speech about AI generation" when using Qwen3 TTS or ChatterBox. github.com/deepbeepmeep/W…

@ilostmyx it doesnt matter how big the gemma file use because when you use memory profile 4 (offloading) its memory footprint is very small. what may take some room in VRAM in the kv cache but using gguf is not going to reduce it.

@deepbeepmeep i was speaking about the gemma file

@ilostmyx both gguf and nvfp4 are available for ltx2. as regards ltx v1, you can use a gguf around and declare a finetune json file (check the doc)

@deepbeepmeep how about possibility to add gguf and quantized models for ltx? because 32gb is too hard...

@JBabaTalks both

@deepbeepmeep Is this old or new released?

WanGP 10.42: More goodies on the way - Flux 2 Klein 4B & 9B: in tiny models get the best of both Qwen Edit & Z Image - LTX2 can now extract both video and audio from a control video - Models Manager: identify easily which files to delete among your 3TB of checkpoints - GGUF & NVFP4 support for lighter models in RAM - RAM optimizations so that you can generate multi minutes videos github.com/deepbeepmeep/W…