SteerTTS: 23M-param English TTS you steer with numbers (pitch, rate, loudness, etc. as z-scores), runs real-time on CPU [P]

Explain Like I'm Normal

SteerTTS is a tiny 23M-parameter text-to-speech model that allows precise control over pitch, rate, and emotion using numerical z-scores. Unlike massive proprietary models, it runs in real-time on standard CPUs and uses a bidirectional Mamba architecture for high efficiency. This gives developers a way to generate expressive, low-latency AI voices without a massive GPU bill.