@nate_512Fishaudio's s2.1 pro is an open-weight tts model that's 2x faster than cartesia and costs 1/6th of elevenlabs. 56.3 chars/s throughput, 83+ languages, ~90ms latency, on-prem with zero data retention. elevenlabs had no real open-source competition until now
عرض المنشور الأصليصورة بواسطة @nate_512· Jul 29, 2026· ElevenLabs
The image is a bar chart displaying "BT Scores - Overall". The focus is on the performance of different text-to-speech models, with "Fish Audio S2 Pro" scoring the highest at 3.07. The chart is clean and data-driven, with a professional aesthetic. A notable detail is the small text at the bottom right stating "5,153 preference pairs min 2 plays each side".
@nate_512Fishaudio's s2.1 pro is an open-weight tts model that's 2x faster than cartesia and costs 1/6th of elevenlabs. 56.3 chars/s throughput, 83+ languages, ~90ms latency, on-prem with zero data retention. elevenlabs had no real open-source competition until now
عرض المنشور الأصليصورة بواسطة @nate_512· Jul 29, 2026· ElevenLabs
The image is a bar chart displaying "BT Scores - Overall". The focus is on the performance of different text-to-speech models, with "Fish Audio S2 Pro" scoring the highest at 3.07. The chart is clean and data-driven, with a professional aesthetic. A notable detail is the small text at the bottom right stating "5,153 preference pairs min 2 plays each side".
