Fish Speech / OpenAudio S1
the code stopped being apache in march 2026 — now neither half of this is open.
genuinely excellent quality behind a licence that has closed further since we last checked: the weights were already non-commercial, and the code stopped being apache-2.0 in march.
we published that fish speech's code was apache-2.0 with only the weights restricted. that was wrong, and it had been wrong for months before we wrote it. the code licence changed to the fish audio research license agreement with the s2 beta commit on 10 march 2026 — a research and non-commercial grant where commercial use requires a separate agreement. the openaudio s1 weights remain cc-by-nc-sa-4.0. both halves of this project are now non-commercial, and github's licence badge is not what tells you so.
that makes it the clearest illustration of this category's central hazard. a repository people have relied on as permissively licensed quietly stopped being one, and nothing about the download experience signals the change. if you built on fish speech before march on the strength of an apache badge, the terms under you have moved.
the description above is ours, condensed from the ranking. pricing moves — check it on the vendor's own page before you rely on it.
- category
- ai text-to-speech models
- pricing
- $11.72 per 1m chars (Pro plan)
- website
- github.com
genuinely excellent quality behind a licence that has closed further since we last checked: the weights were already non-commercial, and the code stopped being apache-2.0 in march.
we researched this category against vendors' own pricing pages and licence files. that is where this line comes from — not from the vendor, and not from anything they paid for.