Community · Oct 22, 2026, 20:08 UTC

AuroraTTS-100K, an English text-to-speech model under 100K parameters, is released

The community-released model runs at 72x real time on a CPU and 1120x on a GPU, and quantized to INT8 it takes 127KB. It can slur difficult words and long numbers.

• ModelAuroraTTS-100K
• Topicaudio
• Topicrelease
• Topicopen-weights