Inflect-Micro-v2: Complete Text-to-Speech Synthesis in Just 9.36M Parameters

What happened

Researchers published Inflect-Micro-v2 on Hugging Face on July 25 -- a text-to-speech model handling the complete TTS pipeline including phoneme conversion, pitch duration, and acoustic decoding in just 9.36M parameters.

Context and impact

Most modern TTS models (ElevenLabs, Kokoro-82M, F5-TTS) have tens to hundreds of millions of parameters. Inflect-Micro-v2 demonstrates viable speech synthesis at a fraction of the compute requirements -- ideal for edge devices, wearables, and embedded systems.

Details

• 9.36M parameters -- compare: Kokoro-82M has ~82M, F5-TTS ~300M
• Complete pipeline: phoneme conversion + pitch duration + acoustic decoder in one model
• Released under permissive license on Hugging Face
• Hacker News score 148 -- strong community interest on July 25