Test: Full music video generated with open models in 1-2 takes
linoy_tsaban · x · 2026-08-25
A creator showcased a video generated entirely using open models, covering both song and visuals, completed in just one or two takes. The workflow involved Music3 for music generation and the H3 model (with audio input support) for video generation. The creator noted the crazy progress of open models in video and music generation, highlighting how significantly they have caught up to closed-source alternatives.
More from Fun
- Critique: Requiring users to be "pure of heart" to use AI is absurd — adamamcbride · 2026-08-25
- Opinion: Rebrand data centers as 'Art & Knowledge Centers' — bennash · 2026-08-25
- Anthropic employee shares surprise company swag — himanshustwts · 2026-08-25
- Ask Codex to beep when it finishes and you get either constant beeping or silence — _akpiper · 2026-08-25
- I put ChatGPT, Claude, and Gemini in a group chat. Please break it. — capibara13 · 2026-08-25
- Copilot plays Wordle with no answer, blames ChatGPT — Dismal-Business-5180 · 2026-08-25