如果要享受 AI 製片的樂趣，這套模型一定要學會

[mtr]

https://www.youtube.com/watch?v=qiT5vJr0I68

[joe]

[mtr]

Sulphur 2 AI 视频生成模型，无审查！8G 显存就能跑 ...

https://x.com/binghe/status/2054601427209912376

[mtr]

Testing out Sulphur2. On the 4090, 1024x1408 goes from 15 seconds to 5 minutes 30 seconds. It feels pretty decent, but there's that characteristic dullness. Whether it's superior to the existing LTX-2.3 is still not entirely clear. This is made by using Dream to stitch together 5 still images.

https://x.com/kiyoshi_shin/status/2053877328166744343

[mtr]

youtube search "sulphur 2"

爆火的「無審查」AI 視訊模型來了！ Sulphur 2 本地部署實測：8G 顯存也能跑！完全免費開源 | 零度解說

https://www.youtube.com/watch?v=Nms56KhvIQw

[mtr]

Build Your Own UNCENSORED AI Running Completely Offline

In this video, I show you how to run a fully unrestricted and private, offline LLM directly on your own machine. If you are a cybersecurity professional, Red Teamer, or Pentester, you know the pain of getting the "I cannot help with that" error when working on legitimate security simulations or code analysis. I also show you how to set up a full private AI with RAG!

https://www.youtube.com/watch?v=SkEUXfv9sEo

[joe]

[mtr]

實操教學：

I made a ComfyUI workflow for running LTX2.3 on an 8GB VRAM setup.

The workflow was tested on an older gaming PC with an RTX 3060 Ti, because I noticed that many people assume LTX video generation is only possible on very high-end GPUs.

The goal is not to push maximum resolution in one pass, but to make the process more stable for low VRAM users.


https://www.reddit.com/r/StableDiffusio ... _workflow/

[joe]