Pika Speech TTS Model Launch

Pika Labs announced Pika Speech, a 3B parameter text-to-speech model with a real-time factor of 0.02. In locally run long-form tests, it generates one minute of studio-quality 48 kHz speech in about 1.2 seconds and supports requests up to five minutes. The model is up to 9× more cost-efficient than ElevenLabs v3, 4.5× than Cartesia and ElevenLabs Turbo, and 2× than Fish Audio.
Key insights
Pika Speech offers significant cost and speed advantages over existing TTS services
Pika Labs claims its new 3B TTS model achieves a real-time factor of 0.02, generating one minute of 48 kHz speech in 1.2 seconds locally, with support for up to five-minute requests. It is 9× cheaper than ElevenLabs v3, 4.5× cheaper than Cartesia and ElevenLabs Turbo, and 2× cheaper than Fish Audio. This positions it as a highly efficient option for long-form speech generation.