VoiceForge Studio
Clone any voice and synthesize authentic speech in seconds using local Qwen3-TTS neural audio nodes running directly on your ComfyUI workstation.
1
Reference Voice Sample
Provide 3 to 15 seconds of clear, uninterrupted speaking
Drop audio file here or browse
Supports WAV, MP3, FLAC, OGG, M4A, or WEBM (5–20s optimal)
00:00
Click the microphone to start recording your voice.
⚡
LANfactory Voice Preset
Clean high-quality reference audio: "Hello from LAN factory! The audio synthesis system is online."
Voice Library Presets:
💡 Tip: If you leave this blank, Qwen3-TTS uses instant zero-shot timbre cloning. If you fill it in, it uses full In-Context Learning for maximum accent fidelity!
2
What Should The Cloned Voice Say?
Type your script or click an inspiration phrase below
Quick Inspiration Scripts:
✨ Emotion Engine: Acting instructions require the 1.7B model and Zero-Shot Timbre mode (leaving reference transcript blank). In ICL mode, the neural network locks onto the reference clip's original prosody.
3
Synthesis Parameters
Fine-tune model weights, language, and creativity
Temperature
0.95
Repetition Penalty
1.05
Voice Output Studio
Synthesized neural speech playback and audio export
Ready for Synthesis
No Speech Synthesized Yet
Load a voice reference sample, write what you want it to say, and hit "Synthesize Cloned Speech" to generate voice audio.
Model
0.6B
Duration
--
Speed
--
Recent Session Takes
Generated audio takes from this session will appear here for comparison.