Stability AI releases Stable Audio Open, a text-to-audio model that generates up to 47 seconds of samples and sound effects, prohibited for commercial use
Later coverage points to two routes for the same technology family: deployment on Arm-based mobile devices and audio models explicitly trained on licensed data. That makes the noncommercial restriction an important boundary between broad experimentation and production use.
First-order effects
Developers and creators can generate audio samples and sound effects up to 47 seconds long with Stable Audio Open, but cannot use those outputs commercially under the stated restriction.
Stability AI expands access to its audio-generation tooling while withholding a commercial-use path for this particular release.
Second-order effects
Teams seeking to place generated audio in products, campaigns, or other revenue-generating work must use another tool or secure a different rights arrangement, rather than rely on Stable Audio Open alone.
The restriction distinguishes experimentation-oriented model distribution from Stability AI’s other audio offerings, including its earlier copyright-free-audio product line.
Third-order effects
If this product segmentation persists, audio AI vendors will increasingly compete on provenance, licensing terms, and deployability—not only clip length or output quality.
The later move to licensed-data Stable Audio 3.0 models suggests commercial audio generation may consolidate around clearer training-data and usage-rights assurances, while more open releases remain constrained.
The trend: Generative-audio vendors are separating widely accessible models from commercially usable systems as rights and deployment requirements become central product features.
Wow, groundbreaking news! Stability AI just unleashed a sound generator that will blow your mind. It's trained on royalty-free recordings and can create amazing sounds and songs. Check it out here: https://techcrunch.com/...
@simonw I'm not a fan of HN's repeated insistance that “open source only means 100% reproducible”, but “open” should mean that the model can be used without hitting roadblocks, with noncommercial use only being a giant roadblock.
We're excited to announce Stable Audio Open, an open source model optimised for generating short audio samples, sound effects and production elements using text prompts. This release marks a key milestone as we further open portions of our generative audio capabilities to [image]
I'm a big fan of Stable Audio (use it a lot!) and seems to fly under the radar more than most other popular AI audio generators like Suno. Worth a look as it can do some cool stuff inc. using your own custom audio. Oh yes, its free and trained on open source data! Yay!
I really wish Stability would stop abusing the term “open source” like this - it's under a custom non-commercial license “Openly licensed” or “open weights” are fine alternatives
Non-commercial license. Stability is dead lol. Remember when I said bearish on open source models? It's cuz business models are equally as important as the technical models
Stability release a slimmer version of their Audio model, under a non-commercial license (reminds me of SD3...) Anyhow, this is a new fun toy to play with, so enjoy!
Stability AI releases Stable Audio Open announcement: https://stability.ai/... model weights: https://huggingface.co/... Weights for Stable Audio model trained on audio data from FreeSound and the Free Music Archive, can generate up to 47 seconds of audio. The model specialises i…