Sogni Voice Playground

Demo Console: This is a live demo for testing the API. Voice clones may be cleared periodically. For production use, clone the repo and run on your own Mac hardware. For OpenClaw follow these steps.
Record voice → get text  ·  Type text → hear it spoken  ·  Clone a voice → speak as anyone
Sogni Voice

API Authentication

1. Voice to Text

Transcribe your microphone live with Parakeet, or record/upload audio for any recognition model.

Parakeet is the fast default for 25 European languages.

Record or upload, then transcribe

Ready to record
Drop an audio or video file here, or use Upload Audio File. Max 25MB. Allowed: mp3, wav, m4a, mp4, webm, ogg, flac. (Video uploads transcribe the audio track.)
No transcript yet. Record audio and it will appear here.
Forced alignment lab Qwen3

Match a known transcript to the audio above. Supports audio up to 5 minutes in 11 languages and returns word timings.

Choose audio above and enter its transcript.

2. Text to Speech

Enter text and receive a voice recording you can play.

Describe how the voice should sound - emotion, accent, speaking style, etc.
Ready to generate
No audio generated yet.

3. Voice Cloning

Record a sample of your voice to create a voice clone that can be used for text-to-speech with any model.

Create New Voice Clone

No transcript needed - just provide a voice sample.

or
Ready to record or upload voice sample
Record or upload a voice sample to create a clone.

Your Pocket TTS Clones

No voice clones yet. Create one above!

Import Voice Clone

Import a previously exported voice clone (.zip file).

Select a ZIP file to import

Create New Voice Clone

Requires a transcript of the voice sample (auto-detected).

or
Ready to record or upload voice sample (3-10 seconds recommended)
Record or upload a voice sample (3-10 seconds recommended).

Your Qwen3 TTS Clones

No voice clones yet. Create one above!

Import Voice Clone

Import a previously exported Qwen3 voice clone (.zip file).

Select a ZIP file to import

Create MOSS Reference Voice

Use 1-30 seconds of clean speech (5-10 seconds recommended). No transcript is needed.

or
Ready to record or upload a voice sample
Record or upload a voice sample to create a reference profile.

Your MOSS Reference Voices

No reference voices yet. Create one above!

Create Fish S2 Voice Clone

Zero-shot cloning from 1-30 seconds of clean speech (5-15s recommended) plus its transcript.

or
Ready to record or upload a voice sample
Record or upload a voice sample and enter its transcript to create a clone.

Your Fish S2 Voice Clones

No voice clones yet. Create one above!