v1.0 is now Open Source

Clone Any Voice.
Speak Any Words.

A privacy-first, CPU-optimized voice cloning engine.
Run it locally. No GPU required. No limits.

Spectral Capture

Instant Voice Cloning

Drop in a 30-second audio sample. Our scanning engine extracts timbre, cadence, and resonance into a lightweight model file. Zero training time.

30s
Min Sample
Local
Inference
INPUT
OUTPUT

Latency? Non-existent.

Our streaming architecture generates audio chunks in real-time. Experience sub-200ms latency on standard consumer hardware.

ArchitectureFlow Matching
Sample Rate24kHz / 48kHz
LanguageEnglish Only
LicenseMIT Open Source
Multi-Speaker

Podcast Studio

Orchestrate complex conversations. Assign unique voices to speakers and generate full episodes with natural turn-taking.

FormatScreenplay Script
Turn-TakingAutomatic
ExportWAV / MP3 / OGG
SpeakersUnlimited Casting
Recording
S
Speaker 1
shreya
A
Speaker 2
ayush

Everything you need.

A complete suite for local audio generation.

Voice Cloning

Clone unique voices from just 30 seconds of audio. Captures nuance, timbre, and cadence.

Natural TTS

Generate speech that sounds human. State-of-the-art flow matching model architecture.

Podcast Studio

Create multi-speaker conversations with a screenplay-style script editor.

CPU Optimized

Runs 100% locally on CPU. No expensive GPUs required for inference.

Multi-Format

Export to WAV, MP3, FLAC, M4A, OGG, and WebM at 24kHz studio quality.

Open Source

MIT Licensed. Transparent, privacy-first, and community driven. Built with Python & Next.js.

Start cloning in seconds.

Get up and running with a few simple commands.

bash — 80x24
$git clone https://github.com/Ayushpani/voiceforge.git
Roadmap Active

Building the future of open voice.

Real-time Streaming GenerationLoading...
Emotion & Style ControlLoading...
Multi-Language Support (v2)Loading...
Voice Marketplace IntegrationLoading...
Mobile App CompanionLoading...

This is an open source project. We need your help to build these features.

Contribute on GitHub