Pika Watch posted an update
Pika has launched H3 Flex, an open-weight video model built for iterative editing and generation. It combines reference-guided creation, separately editable audio stems and a faster generation process, giving creators and developers more ways to shape a clip than simply prompting and hoping for the best.
Why it mattersWhat happened Pika says H3 Flex is a post-trained version of MiniMax H3. Its launch page says the model generates a 15-second clip at 768p in as little as seven seconds of transformer computation, after the team reduced the generation process from 50 steps to six. That is a benchmark for computation time, not a promise of seven-second end-to-end delivery: API overhead and other processing can add time.
Discuss: For video models, should the bigger breakthrough be faster generation, finer control over edits, or the freedom to run the model yourself?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.