A new family of local text-to-speech (TTS) models, named Inflect v3, has been released. The collection includes three variants: Mini (34.31M parameters), Micro (11.79M parameters), and Atom (157K parameters). The Mini model offers advanced features such as vocal-event tags for sounds like laughter or sighs, reference voice-tuning, and 48 kHz enhancement, and can run on both CPU and CUDA. AI
IMPACT Provides developers with new, lightweight options for local text-to-speech synthesis, potentially enabling more efficient AI applications.
RANK_REASON Release of a new open-source model with detailed technical specifications and performance metrics. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →