Pika releases H3 Flex for six-step joint audio-video generation
Pika released Pika H3 Flex, a derivative of MiniMax H3 that combines video and synchronized stereo-audio generation through six denoising steps. The open release includes two LoRA adapters, MiniMax H3 base components, local inference tools, ComfyUI workflows, and Hugging Face Diffusers integration. Pika says it natively supports 480P and 768P output for clips of about 5 to 15 seconds, while its API offers higher-resolution enhancement and separate audio tracks. Why it matters: The release gives developers open adapter weights and inference code for a six-step audio-video model with local, ComfyUI, Diffusers, and API workflows. What do you think?
Open the headline: source, AI briefing and more →Comments
…