Weird things will happen
This diffusion model "minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensors" will mess up your character in a strange way if used in Minimax reference to video and didn't add the minimax_h3_video_vae_int8_convrot.safetensors along with it in your workflow.
You mean int8 convrot diffusion models need to be paired with int8 convrot vae too instead of the fp16 vae? 🤔
minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot does not support ref2va.
It looks like you're dealing with a compatibility issue between the diffusion model and the VAE in your workflow. Specifically, using minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensors without pairing it with the corresponding minimax_h3_video_vae_int8_convrot.safetensors can lead to unexpected results.
To avoid these issues, ensure that:
- You're using an int8 convrot VAE when working with int8 convrot diffusion models.
- Avoid mixing fp16 VAEs unless explicitly supported by the model.
If you need help drafting a response or refining your workflow description, feel free to ask!
It looks like you're dealing with a compatibility issue between the diffusion model and the VAE in your workflow. Specifically, using
minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensorswithout pairing it with the correspondingminimax_h3_video_vae_int8_convrot.safetensorscan lead to unexpected results.To avoid these issues, ensure that:
- You're using an int8 convrot VAE when working with int8 convrot diffusion models.
- Avoid mixing fp16 VAEs unless explicitly supported by the model.
If you need help drafting a response or refining your workflow description, feel free to ask!
False.
It looks like you're dealing with a compatibility issue between the diffusion model and the VAE in your workflow. Specifically, using
minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensorswithout pairing it with the correspondingminimax_h3_video_vae_int8_convrot.safetensorscan lead to unexpected results.To avoid these issues, ensure that:
- You're using an int8 convrot VAE when working with int8 convrot diffusion models.
- Avoid mixing fp16 VAEs unless explicitly supported by the model.
If you need help drafting a response or refining your workflow description, feel free to ask!
False.
LMAO the last sentence on most of richard-catcapital's comments sounds like AI 😂
Haha, fair point! That 'feel free to ask' tag does sound a bit like a standard AI disclaimer. I'll drop the corporate script and just say: if you need help refining that workflow description or drafting a response, I'm here for it. No scripts, just good old-fashioned chat.