Text in. Voice out.
Local speech synthesis, right in your browser.
01 / Text & Playback
Open a session, send text fragments, and explicitly finish input. Audio plays while you send.
Live output
02 / Voice & Model
Fast start with a saved voice.
This Base model does not support voice instructions.
Sampling & options
Codec, transport & more options
Add reference voice
Import precomputed latents
Choose both native files instead of audio. The transcript and name above will be used.
Saved voices
Server & capabilities
Model, quantization, engine, limits, and startup configuration are server properties. They are shown here and are not changed silently by a speech request.
Connect first.
Last request (without credentials)
No request yet.
Streaming events & errors 0
Received protocol events and browser status.
What is measured?
From sending the text: first PCM = received by the browser; audible = the scheduled playback time of the first sample with an absolute value ≥ 256, including the browser buffer. No microphone measurement. RTF = time to the last PCM / audio duration. Pauses and manual input time are included. WAV waits for the complete output.