Loading…

The most interesting things people are building with AI, all in one place — followed and tracked as they evolve from first demo to real product.

Press / to search--:--:-- ET

Dropday is new — we're looking for genuine feedback.

Home Leaderboard Top creators What is this?Privacy Terms

A project by@gruberbuilds

Get the best builds by email. No account needed.

Home Explore Shipped Top Drops

Home Explore Shipped Leaderboard Model Drops

Sign in

H3 Flex — Model Drop — Dropday.ai

Home Explore Shipped Leaderboard Model Drops

Sign in

← All model drops

Live drop

H3 Flex

PikaDropped Oct 29, 2026

6-step distilled video gen with native stereo audio stems and up to 12 multimodal refs for rapid iteration.

Browse in feed Submit a build

About this release

What is H3 Flex?

The most interesting things people are building with AI, all in one place — followed and tracked as they evolve from first demo to real product.

4581 indexedPress / to search--:--:-- ET

Get the best builds by email. No account needed.

Dropday is new — we're looking for genuine feedback.

Home Leaderboard Top creators What is this?Privacy Terms

A project by@gruberbuilds

DROPDAY.ai

Home Explore Shipped Top Drops

Pika H3 Flex is an open-weights post-trained video model distilled from MiniMax H3 to 6 sampling steps. It generates up to 15-second 768p clips with synchronized stereo audio (exportable as 5 separate stems: dialogue, score, SFX, Foley, ambience), supports up to 12 reference inputs (sketch, image, video, voice, moodboard), and offers draft-to-4K upscaling. Weights and inference code are available on Hugging Face and GitHub under the MiniMax H3 Community License; API access via Pika API Club.

Read the official Pika launch page

First builds incoming — this lane fills in automatically as people start building with H3 Flex. Know one? Submit the URL.