Pika released H3 Flex on October 29, introducing an AI video model designed to produce short clips faster while giving editors more control over characters, camera movement and sound. Built by post-training MiniMax H3, the system accepts up to 12 visual, audio and video references and can generate synchronized footage with sound. Pika also published model adapters and inference code for developers who want to run supported versions on their own hardware.

The startup says H3 Flex can produce a 15-second clip at 768p with as little as seven seconds of transformer computation. That is a benchmark of one part of generation, not a guarantee that an uploaded request returns a finished video in seven seconds. Input processing, decoding, networking and other steps can add time.

From initial draft to a finished edit

The model uses six denoising steps instead of the 50 steps of its parent system, according to Pika’s technical report. Creators can supply character references, existing movement and keyframes to steer a shot, or change elements of footage while preserving selected motion and timing. That gives the product a different emphasis from tools designed mainly to create a clip from one text prompt.

In one published example, multiple camera angles and character actions form a connected narrative. The clip is a developer-selected showcase rather than evidence that every complex prompt will produce consistent characters or physically accurate motion.

For audio, H3 Flex can produce synchronized stereo sound. Its hosted API can additionally export five editable stems—dialogue, music score, effects, Foley and ambience—so an editor can adjust one layer without replacing the entire mix. The company offers 480p and 768p native generation; 1080p, 2K and 4K results come from a hosted upscaling process, not native 4K inference.

Open weights, with practical limits

Pika published downloadable model components and inference code for local work. However, access and redistribution are governed by the applicable licenses, and the advanced upscaler, separated audio stems and accelerated hosting remain Pika-managed services.

Pika began offering the system through its API on October 29 and says it plans to bring it into its creation product on November 2. The move intensifies a market already shaped by Kling 4.0, FLUX 3 Video and emerging interactive world models. How much time H3 Flex saves editors in real projects will depend on quality, revision needs and the hardware used.

Editor’s Note: We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support our work and help further secure the integrity of our research initiatives, please scan the QR code below to donate 5,000 sats.

Scan the QR code using a compatible Bitcoin wallet while viewing the article on a laptop or larger screen. Donation address: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb.

BitcoinVersus.tech is not a financial advisor. This media platform reports on financial subjects purely for informational purposes.