Export
The link always opens the latest version of this page.
Ask an AI about it
Open in ChatGPTA new chat that reads this entry from the site Open in ClaudeThe same, in ClaudeShare the link instead of a file, and they will always see the latest version.
Files are a snapshot of today. The live page has anything newer.
Inflect-Micro-v2
Text-to-speech synthesis model with 9.36M parameters, 37.53 MB FP32 footprint, optimized for CPU/CUDA inference
- Category
- text-to-speech synthesis 1 source
Inflect-Micro-v2 is a lightweight text-to-speech synthesis model developed by Owen, optimized for CPU and CUDA inference. The model contains 9,356,513 parameters with a 37.53 MB FP32 footprint 1.
The v2 release achieves a UTMOS22 quality score of 4.395 (95% CI 4.381–4.408) 1, with 66.2% community preference compared to KittenTTS, Piper, and Supertonic 3 1. Semantic accuracy, measured via two-ASR evaluation (Qwen3/Nemotron mean), reaches 3.99% word error rate 1.
Performance on 4-thread CPU systems reaches 6.28× real-time throughput 1. The model outputs 24 kHz mono audio, supports deterministic seeding, and handles long-text input 1. Implementation is built on PyTorch 1.
Facts 2
| Attribute | Value | Confidence | Sources | Cited |
|---|---|---|---|---|
| Vendor | Owen | low | 1 | 1 |
| Category | text-to-speech synthesis | low | 1 | 1 |
Platform 1
Official links
Sources 1
- Inflect-Micro-v2: complete voice in 9.36M parameters · Hacker News (YC)