Home/New releases/Pika H3 Flex

Pika H3 Flex

Pika's open-weight video model with sound, post-trained on MiniMax H3, that renders in six steps and finishes up to 4K.

Pika published Pika H3 Flex on October 29, 2026. It generates video with synchronized sound from text, an image, or references, including multiple image, audio, and video references. Pika also lists editing uses: style transfer, motion transfer, object replacement, character swap, and targeted edits. Pika built it by post-training MiniMax H3. The Hugging Face repo is a LoRA adapter on MiniMaxAI/MiniMax-H3, used with Diffusers. Sampling goes from 50 steps down to 6. It generates natively at 480p and 768p. A separate upsampler finishes to 1080p, 2K, and 4K. Pika returns five editable audio stems: dialogue, score, SFX, foley, and ambience. You can use it through the Pika API, ComfyUI, and Diffusers. Weights are on Hugging Face and code is on GitHub. The weights are open under the MiniMax H3 community license, per the model card. Pika's listed starting price on its API is $0.01 per output second for text or image to video and $0.03 per second for reference to video. Check Pika's site for full pricing. Pika's own claim: it ranks first among open-source text-to-video models. Pika also notes that physics, multi-scene consistency, and complex reference control still need work. Pika's own speed numbers were measured on eight B300 GPUs. For most people the API is the practical way in. Last checked November 2, 2026.

• Runway alternatives
• Image and video list