Not affiliated with Tencent in anyway, Just uploading these here for convenience, and I am a fan of Hunyuan Video.

Some info:

My Setup: 720p T2V FP16 with LightX2V Lora - I run lora at 0.5 with 8 steps, can run 4 steps on low motion - UPC/Beta11
SweetMax797: Runs I2V with LightX2V Lora - 4 steps with Euler/Normal

LightX2V Tested and working (4 or 8 steps)

T2V 720p FP16
T2V 480p FP16
I2V 720p FP16
I2V 480p FP16

LightX2V Tested and NOT working (Full 50 step generations)

480p FP8 CFGD Scalled (Distilled Model)

My full setup is a multi gpu / multi stage setup.
I run low res step first with no lightx2v for motion, then use the HYV 720p latent upscaler into second sampler pushing 720p or higher at 4-8 steps with .75 - .85 de-noise. This works for me because I'm focused on high motion non erotic scenes.
i do run just single pass 720p on low motion though.

Search AI Models & Addons

very few results