Pika Introduces Enact, a Research Preview for Directing AI Actors and the Camera in Real Time
VP Land
Nov 18, 2026 · 2 min read
Pika introduced Enact, a research preview built on its Pika H3 Flex model that lets a creator interact with AI "actors," re-stage a scene and move the camera in real time, according to the company's announcement on X. Blocking and camera changes happen inside a scene that responds to the creator, an approach Pika says is "designed to mimic the experience of directing on set."
Built on Pika H3 Flex. Enact sits on top of Pika H3 Flex, the company's post-trained version of MiniMax's H3 model.
Demos on request. Pika asks interested creators to reach out for a demo. The announcement lists no pricing, release date or public access.
Open model underneath. Pika H3 Flex has open weights, but the announcement does not say Enact itself is open.
Key specs unstated. The announcement does not give Enact's output resolution, scene length or latency.
Enact lets creators re-stage a scene and move the camera while it responds live
Pika describes Enact as "a research preview of scenes that respond to you." The announcement does not say whether Enact will come to the company's main creative platform, which generates video, images and audio.
The three controls Pika names map to familiar on-set jobs: working with performers, changing the blocking, and moving the camera. Pika calls Enact "a new way of making AI video."
A frame from the company's demo shows a subtitled line from a character named Alice.
A frame from Pika's Enact research preview, with a character line reading "ALICE: Yeah, we're ready." Image: X (formerly Twitter)
The announcement leaves several practical questions open. It does not say how creators interact with the actors, how many characters a scene supports, how long a scene can run, what resolution Enact outputs, or how much latency "real time" means in practice.
Pika H3 Flex supplies the model underneath, post-trained from MiniMax H3
Enact's foundation is Pika H3 Flex, which Pika describes as its post-trained MiniMax H3 model, built for creative iteration. Pika lists support for up to 12 references, five separate audio stems, finishing at up to 4K, and open weights.
Pika's technical deep dive describes H3 Flex as an omni-generator that produces synchronized video and stereo audio of up to 15 seconds in six denoising steps. Pika says that is about 8x fewer steps than its 50-step teacher model.
The announcement does not say which of these H3 Flex capabilities carry into Enact.
Pika is offering Enact demos on request while H3 Flex weights are public
The post closes by inviting creators to "reach out for a demo," and Pika links to a "Research Preview: Pika Enact" page on its site for more detail. The announcement gives no pricing or release timing.
The model underneath is easier to get. H3 Flex's open weights are published on Pika's Hugging Face page, where the listing describes it as a 33B-parameter image-to-video model.
Those open weights cover H3 Flex. Nothing in the announcement indicates that Enact's real-time directing layer is open as well.
Apple released FastFrames on Hugging Face, two autoregressive video diffusion models that generate 480p video at interactive latency, with the 1.3B FastFrames-S producing a frame every 36 ms on one B200 GPU and the 14…
Comfy added custom Builds to Comfy Cloud, letting a team package its own ComfyUI nodes, models, LoRAs and private node packs into one setup that every artist opens in the browser, as the company said on X. Studios now…
Krea launched Video Swap, a new workflow in Krea Video that replaces people, outfits, products, objects or backgrounds in existing footage using up to four reference images. A team can re-cast, re-dress or relocate a…