Hugging Face's logo

Comfy-Org
/
MiniMax-H3

Diffusion Single File
comfyui
Model card Files
xet
 Community
74

Workflow links: Pika H3 Flex templates

#71
by ComfyUI-Wiki - opened
base: refs/heads/main
←
from: refs/pr/71
Discussion Files changed
+9
-2
ComfyUI-Wiki
 Comfy Org org

3 templates from Comfy-Org/workflow_templates that run the Pika H3 Flex adapters on MiniMax H3: Text to Video, Reference to Video, and FL2VA.

  • Rows appended to the existing ## Workflows table, 200 px single-preview thumbnails, ordered Text to Video, Reference to Video, FL2VA.
  • Weights: loras/pika_h3_flex_fl2va_v1_bf16.safetensors and loras/pika_h3_flex_ref2va_v1_bf16.safetensors, added in PR #70.
  • Tutorial page: Comfy-Org/docs#2175.

Nothing else in the README is changed.

This change was prepared with AI assistance (ComfyUI-Wiki) and reviewed before merge.
Workflow links: Pika H3 Flex templates3c6aba03
supermind
•
edited 7 days ago

Does pika h3 flex work with pruned bf16 models? I got bad results
photo_2026-10-31_22-28-01
Workflow links: add missing templates13f6b4b7
 Workflow links: add missing templatesc5190abf
windowsex

Does pika h3 flex work with pruned bf16 models? I got bad results
photo_2026-10-31_22-28-01

Could you double-check your million-resolution latent block/tile setting over there? Is it set to 32 or something else? Setting it to 32 is super important for proper decoding and alignment—otherwise, it tends to mess things up.
Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment