# [AI Video] MiniMax-H3 All-in-One Video Generation V4 Workflow - RH Edition
# [Features] Latest acceleration scheme; Official skill for automatic/manual prompt optimization; Text-to-video, image-to-video, and reference-to-video three-in-one; Latent tiled secondary sampling; Small faces are non-greasy and stable; Digital human audio can be locked unchanged; Clean management of image, video, and audio assets
# [Author] Astral 星芒
# [Workflow Tutorial]: https://space.bilibili.com/176339505
# [Online Workflow]: https://www.runninghub.ai/post/2097308979525799937/?inviteCode=rh-v1595
# [Local Workflow + Models]: https://pan.quark.cn/s/f8af7c9e64f1
# [Local Workflow + Models备份]: https://pan.baidu.com/s/1mGU37ycAzP0xV0RG9fdySg?pwd=htr9
Description
FAQ
Comments (2)
There is a part that needs to be modified in the ContextIR node. In the h3_tiled_sampler.py file, class H3TiledSampler(io.ComfyNode):
@classmethod
@classmethod
def define_schema(cls):
return io.Schema(
At this point, the single line @classmethod should be removed.
Thank you for your feedback. It should have been fixed by now.