Pika Research

We build video and audio models for creative expression, turning frontier research into new ways to create.

Explore Pika

Website · X / Twitter · Github · Developer API

Featured release: Pika H3 Flex

Pika H3 Flex generates video with synchronized audio, with support for text prompts, first/last-frame conditioning, and multimodal references. It uses six neural function evaluations (NFEs) for generation.

• Try the demo
• Model weights
• Inference code and quickstart
• Technical deep dive
• ComfyUI tutorial
• Diffusers documentation

Contact

For inquiries, reach us at contact@pika.art.