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.