quants for LTX video, the original was fp32 but was being loaded as bf16 so this is supposed to not make any difference

this is realllly fast! below is logs on my 8GB laptop!

100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 30/30 [01:06<00:00, 2.22s/it] Prompt executed in 71.85 seconds

To run this in ComfyUI install the wrapper using comfy manager https://github.com/Lightricks/ComfyUI-LTXVideo

Output matching original results while more stable and faster: image/gif

Example 2: Input image image/png) Prompt

The toy stands on a light blue surface, dressed in a black suit with a red tie. It has a smooth, round face with large eyes, a small nose, and rosy cheeks. The toy begins to move by shifting its weight from one foot to the other. It raises its right arm, then lowers it while lifting its left arm.The toy's feet tap in place as it moves. It starts rotating in small, smooth circles, its arms swinging with each spin.The toy performs several spins, its feet slightly lifting off the ground with each turn, before landing gently back on the surface.The toy's body and limbs move in rhythmic, fluid motions, with its head following the movement of its arms and body. Output:

Tip: for I2V you should introduce CRF to add a bit of mp4 compression noise, use the workflow provided

Working on more quants!

feedback is welcome.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Unable to determine this model's library. Check the docs .

Model tree for MayensGuds/ltx-video-quants

Finetuned
(9)
this model