My H3 Settings that generate 15 second videos in 2 mins.
retained text
**EDIT: Workflow and sample added to comments.** A quick post for anyone who is struggling with getting good results at a good gen speed. The model can be found [here](https://huggingface.co/OliviaRossi/Minimax-h3_Singularity-x-10Eros-Max_PRISM/blob/main/10Eros_Max_SingularityV2_PikaFlex_0.85_int8HQ.safetensors) Veda Sparse Attention Node can be found [here](https://github.com/veda-sparse/Veda-on-ComfyUI.git) While the video quality is great! What surprises me most about this is the audio quality. It remains very high. I want to be very, very clear here. These settings are not for maximum realistic showcase videos. This is to get well-made videos at a good speed. So you don't need to jump into the comments talking about how shit this all is. As always, your mileage may vary. Specs: RTX 5090. 64 GB RAM. Edit: Sorry, no workflow available. But it's easy to add these into any workflow you may be using.
document fetched 2026-11-03 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xly4rw/my_h3_settings_that_generate_15_second_videos_in