TaoMate H3 FL2VA checkpoint converted for ComfyUI at half the size
robomar_ai_art · reddit · 2026-10-15
A Reddit user converted TaoLive's new TaoMate H3 FL2VA checkpoint to ComfyUI format and uploaded it to Hugging Face. The BF16 file is 1.24GB (down from 2.48GB), requiring no retraining or merging—just a precision-reduced format conversion. It targets 3-step generation; the official streaming and KV caching features aren't included, and first/last-frame support depends on the workflow.
More from Multimodal
- MiniMax-H3 hailed as favorite open release of the year, runs fully local — AdinaYakup · 2026-10-23
- Michal Malewicz experiments with depth maps in his creative workflow — michalmalewicz · 2026-10-23
- One-Shot Tunable Style Transfer on Flux3 Via Reference Latent + Node — Winter_unmuted · 2026-10-23
- PhysVidBench accepted at EMNLP 2026: do video gen models grasp physical commonsense? — aykuterdemml · 2026-10-23
- EMNLP 2026 paper teaches multimodal LLMs to get cartoons via incongruity-resolution supervision — aykuterdemml · 2026-10-23
- A gorgeous Midjourney style: sref code 820221253 for dreamy lavender watercolor looks — umesh_ai · 2026-10-23