Readme

NeuTTS Nano

English text-to-speech with instant voice cloning, powered by neuphonic/neutts-nano.

Ultra-compact (~117M active params) on-device TTS that clones a speaker from a few seconds of reference audio and synthesizes natural English speech at 24 kHz.

Examples

Preset voice

Input - text: My name is Andy. I'm 25 and I just moved to London. The underground is pretty confusing, but it gets me around in no time at all. - voice: jo

Output: 24 kHz WAV speech in the Jo reference voice.

Custom voice clone

Input - text: This is a quick demo of NeuTTS Nano voice cloning on Replicate. - ref_audio: a clean 3–15s mono WAV of the target speaker - ref_text: exact transcript of that reference clip

Output: WAV speech matching the uploaded speaker.

Inputs

Name	Type	Default	Description
`text`	string	required	English text to synthesize
`voice`	choice	`jo`	Built-in reference: `jo`, `dave`, `emily`, `paul`, `sophie`, `steven`
`ref_audio`	file	optional	Custom 3–15s reference WAV (overrides `voice`)
`ref_text`	string	`""`	Transcript of `ref_audio` (required for custom cloning)
`temperature`	float	`1.0`	Sampling temperature
`top_k`	int	`50`	Top-k sampling cutoff
`seed`	int	`-1`	Seed (`-1` = random)

Tips

Prefer clean, continuous speech references (mono, 16–44 kHz, 3–15 seconds).
For best cloning quality, keep ref_text faithful to the reference audio.
Outputs are watermarked by NeuTTS / Perth by default.