LaunchVisual AIOctober 30, 2026

Pika releases open-source H3 Flex video model

Read original source →huggingface.co

Pika H3 Flex is a post-trained MiniMax H3 that generates in six denoising steps instead of 50, priced from $0.01/sec and ranked #1 among open-weight text-to-video models on DesignArena. It outputs up to 4K with five editable audio stems (dialogue, score, SFX, foley, ambience) — a first for a video model.

How this story unfolded

2 days · 1 report · 8 community posts · from Oct 29

  1. Oct 29
  2. Oct 30
  3. Oct 31

More stories today

Open the live feed