https://github.com/denofequity/forge_space_ltxvideo

LTXvideo text-to-video and image-to-video as a Space in Forge2
https://github.com/denofequity/forge_space_ltxvideo

image-to-video ltxvideo text-to-video

Last synced: 8 months ago
JSON representation

LTXvideo text-to-video and image-to-video as a Space in Forge2

• Host: GitHub
• URL: https://github.com/denofequity/forge_space_ltxvideo
• Owner: DenOfEquity
• Created: 2024-12-29T21:28:47.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2025-01-03T21:15:07.000Z (11 months ago)
• Last Synced: 2025-02-04T18:42:53.359Z (10 months ago)
• Topics: image-to-video, ltxvideo, text-to-video
• Language: Python
• Homepage:
• Size: 43.9 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
## Forge2 Spaces implementation of LTX-video ##

New Forge only.

* https://huggingface.co/Lightricks

* https://github.com/Lightricks/LTX-Video

* https://huggingface.co/spaces/Lightricks/LTX-Video-Playground

uses *diffusers*

will download ~10.8GB models

>[!NOTE]

>Install via *Extensions* tab; *Install from URL* sub-tab; use URL of this repo.

>

>Requirements:

>`diffusers>=0.32.0`

>

>`imageio-ffmpeg`

>[!TIP]

>`noUnload` toggle keeps models in memory

>

>`lowVRAM` one-way switch enables sequential model offloading - much lower VRAM usage, but slower: usually better than hitting shared memory.

>[!IMPORTANT]

>Size / frames you can generate is dependent on available VRAM. Start low. Most likely fail point is VAE, after inference.

>

>Results saved to standard *Outputs* folder. Filename is *LTXvideo_**datetime**.mp4*.

>[!NOTE]

>Also saves a text file with same name (*LTXvideo_**datetime**.txt*), containing generation parameters.

https://github.com/user-attachments/assets/4da0bf8a-63aa-4b21-85b0-d0af21ff095a