Start with a spark
Write a scene, upload keyframes, or bring a reference. Tell the model what happens and what we hear.
Turn words, images and references into moving pictures with sound. Make your first scene in seconds of footage, not hours of editing.

Explore scenes from text, images and references, including official Pika H3 Flex examples.
Write a scene, upload keyframes, or bring a reference. Tell the model what happens and what we hear.
Choose your duration, aspect ratio and resolution. See the credit cost before you create.
Preview your scene, upgrade a draft to 4K, and download the result when it feels right.
Create short scenes with synchronized audio from text, keyframe images or reference media. Use Draft to Video to upscale a completed 480p or 768p scene to 1080p, 2K or 4K.
Text and image modes cost 2 / 4 / 7 / 8 / 16 credits per second at 480p / 768p / 1080p / 2K / 4K. Prompt refinement adds 4 credits. Reference mode uses a higher rate, with a fixed 15-second input allowance when video references are included. The full price is shown before submission.
A confirmed provider failure restores the credits automatically. If a request times out, it remains saved so you can refresh My videos without paying for the same job twice.
Yes. Image mode accepts a first frame, last frame or both. Reference mode accepts images, videos and audio; include at least one image or video. Only upload materials you have permission to use.
480p and 768p are native resolutions. Higher resolutions use an upscale stage that preserves the scene and audio. Video duration is 5–15 seconds.