Plachataa Web

1. Source audio

The voice you want to change

Drop an audio file here, or

Choose file

Recent uploads

2. Reference voice

Who it should sound like (25 s max used)

Drop an audio file here, or

Choose file

Recent uploads

Voice library

No saved voices yet. Pick a reference and click “save to library”.

Real-time voice changer

Live microphone → reference voice

Uses the reference picked above and the small real-time model (~0.5 s delay on an RTX card). Wear headphones. To use it in games or calls, choose a virtual audio cable as output.

Microphone Output

Latency preset:

Advanced settings

Block time (s) Audio per step. Shorter = less delay but needs a faster GPU. Diffusion steps Right context (s) Look-ahead. More = better quality, adds delay 1:1. Left context, encoder (s) Left context, DiT (s) Crossfade (s) CFG rate Reference prompt length (s) Noise gate (dB) -60 = off. Blocks quieter than this are muted.

Play converted voice here

3. Model & settings

Model

Presets:

Diffusion steps Higher = better quality, slower. 10 is fine for speech, 30-50 for singing. Length adjust <1 faster speech, >1 slower speech.

CFG rate

Auto pitch adjust (shift source pitch to the reference’s range) Pitch shift (semitones)

Intelligibility CFG Similarity CFG Top-p Temperature Repetition penalty Convert style / emotion / accent too Anonymise only (ignore reference, use an average voice)

4. Results

Nothing converted yet.