Pika releases Pika H3 Flex open weights for video and audio
Original title: Pika-Labs/Pika-H3-Flex — new model trending #29 on Hugging Face
Local inference outputs native 480p or 768p with stereo audio; the hosted API adds up to 4K.
Pika Labs published Pika H3 Flex, a post-trained model built on MiniMax-H3, now trending at number 29 on Hugging Face. It generates video with synchronized stereo audio in six sampling steps and accepts up to 12 reference files through Ref2VA. Local weights cover 480p and 768p at 24 fps for requested lengths of 5 to 15 seconds; the Pika API adds 1080p, 2K and 4K plus audio stems.
Unconfirmed: The model card says showcase clips are not guaranteed to reproduce with the released package.
- FL2VA covers text-to-video and first- or last-frame conditioning.
- Ref2VA accepts up to 9 images, 3 videos and 3 audio clips.
- Output is a 24 fps MP4 with 32 kHz stereo AAC audio.
- A 5-second request produces 124 frames, about 5.17 seconds.
- Vendors
- Pika Labs
- Products
- Pika API
- Organizations
- Hugging Face
- AI models
- Pika H3 FlexMiniMax-H3
From the original article
6 min read at huggingface.co<img src="assets/pika-h3-flex-banner. png" width="1000" alt="Pika H3 Flex" /> <p align="center"><strong>Built for professional iteration. </strong></p> <p align="center">Six-step generation · Omni-reference · Native 480p / 768p · Up to 4K via Pika API</p>
By Pika-Labs · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.