Stable Audio's open music model drops

PLUS: Gemini Omni drops free on YouTube Shorts, and OpenAI solves an 80-year geometry puzzle

Author

Gen AI Art
May 22, 2026 • Estimated Reading Time: 3 minutes

In partnership with

Stability AI just released Stable Audio 3.0 — a family of open-weight music models that generates full songs over 6 minutes long, trained entirely on licensed audio. The small and medium models are free to download, fine-tune, and use commercially.

For creators who’ve been cautious about AI music tools over copyright concerns, this changes the equation: a commercially safe, open foundation model you can train on your own sound library. Is this the moment AI music production finally goes mainstream?

Today in AI:

  • Stable Audio 3.0 drops with open weights and six-minute song generation

  • Gemini Omni converts any input — image, audio, text — into physics-aware video

  • OpenAI’s reasoning model disproves an 80-year-old geometry conjecture

The IT strategy every team needs for 2026

2026 will redefine IT as a strategic driver of global growth. Automation, AI-driven support, unified platforms, and zero-trust security are becoming standard, especially for distributed teams. This toolkit helps IT and HR leaders assess readiness, define goals, and build a scalable, audit-ready IT strategy for the year ahead. Learn what’s changing and how to prepare.

What’s new? Stability AI released Stable Audio 3.0, a family of four audio generation models — including an open-weight medium model capable of producing songs over 6 minutes long — trained exclusively on licensed music and free for commercial use.

What matters?

  • The medium and large models generate over 6 minutes of audio with sustained musical structure and melodic coherence, plus support for audio inpainting and causal continuation — letting creators extend compositions beyond the initial generation.

  • Open weights for the small SFX, small, and medium models are available on GitHub with LoRA fine-tuning documentation, so developers can train custom models on their own audio libraries.

  • Unlike most open-weight audio tools, all outputs carry full commercial rights under the community license — Stability AI trained only on licensed tracks, with partnerships including Universal Music Group and Warner Music Group.

Why it matters?

Copyright risk has been the biggest blocker for developers building on AI music tools. Stable Audio 3.0 removes that barrier with a licensed, commercially safe foundation model that’s open enough to customize.

GUIDE

What’s new? Google launched Gemini Omni at I/O 2026 — an any-to-any model that takes text, images, audio, or existing video as input and outputs physics-aware, conversationally editable video. Omni Flash, the first model in the family, is live now and free on YouTube Shorts.

What matters?

  • Omni reasons across all input types to produce coherent video — a prompt like “a claymation explainer of protein folding” generates a stop-motion video with voice-over in a single pass.

  • Conversational editing lets you describe changes in plain language after generation — “shift the camera angle to the left” or “make it ripple like liquid when touched” — without re-generating from scratch.

  • All Omni-generated videos include SynthID watermarking for provenance tracking, with an identity-verification step required to use your own avatar in order to prevent deepfakes.

Why it matters?

Free access on YouTube Shorts puts AI video creation in front of millions of creators who never used a dedicated AI video tool. Conversational editing removes the biggest frustration in AI video workflows — having to start over when you want to change just one thing.

SPONSORED BY MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

What’s new? OpenAI announced that an unreleased general reasoning model independently disproved Paul Erdős’ 1946 planar unit distance conjecture — a long-standing open problem in discrete geometry — verified by Fields Medalist Tim Gowers and Princeton mathematician Will Sawin.

What matters?

  • Erdős’ problem asks how many point pairs in a flat grid can sit exactly one unit apart; for 80 years, square-grid arrangements were considered optimal. The model found an entirely new infinite family of point configurations that exceeds the grid, approaching it through algebraic number theory rather than iterating on known solutions.

  • Unlike DeepMind’s AlphaProof, which was purpose-built for formal mathematics, this breakthrough came from a general-purpose reasoning model not yet released to the public.

  • OpenAI researcher Alex Wei called math “a leading indicator of what is to come,” framing this as a signal that AI is moving toward “Level 4” capability — making independent discoveries rather than accelerating existing research.

Why it matters?

AI independently disproving an open mathematical conjecture marks a different kind of capability than solving known problems. If this generalizes, the question becomes which unsolved problems across science, design, and creative research are next on the list.

Everything else in AI

Odyssey launched Agora-1, the first multi-agent world model, letting up to four players simultaneously experience a GoldenEye-style deathmatch being generated frame-by-frame in real time — each participant sees the same AI-generated world from their own independent viewpoint.

Anthropic hired OpenAI co-founder Andrej Karpathy to lead a new research unit focused on using Claude to accelerate the pretraining process, marking one of the biggest AI talent moves of the year.

Samsung and Google revealed Android XR smart glasses at I/O 2026, partnering with Warby Parker and Gentle Monster for a fall 2026 launch with Gemini-powered live translation, navigation, and real-time contextual awareness.

Essential AI Guides - Reading List:

Let us know!

Work with us

Reach 100k+ engaged Tech Professionals, Engineers, Managers and decision makers. Join brands like MorningBrew, HubSpot, Prezi, Nike, Ahref, Roku, 1440, Superhuman, and others in showcasing your product to our audience. Get in touch now →