Pika releases open-weights H3 Flex video model

Pika has introduced Pika H3 Flex, its post-trained version of MiniMax's H3 video model, and is releasing the generator adapters and inference code so developers can run it on their own hardware under the model's license terms.

The main technical change is speed. Pika says it post-trained H3 to generate in six steps instead of 50, then refined the faster model for quality and control. The company says Flex can produce a 15-second 768p clip with synchronized audio in as little as seven seconds of transformer (DiT) computation. That figure comes from a controlled benchmark and does not reflect total end-to-end API latency. At launch, Flex ranked #1 among open-weights models and #7 overall on DesignArena's text-to-video leaderboard, according to Pika.

Flex accepts up to 12 reference files, mixing images, video and audio, in a single generation. It also supports reference-guided editing, such as swapping a prop or outfit while keeping the performance and timing. It generates natively at 480p or 768p with synchronized stereo audio.

The open release covers text-to-video, first- and last-frame generation and multimodal reference-to-video at 480p and 768p. Several features stay hosted by Pika: upscaling to 1080p, 2K or 4K, separation of audio into five stems (dialogue, score, SFX, Foley and ambience), optional prompt enhancement and the accelerated serving stack.

Weights are on Hugging Face and code is on GitHub. Flex is available now through the Pika API and was scheduled to reach Pika's creation product on November 2. Pika did not give pricing or details of the license terms in the announcement.