Glossary
Voice, audio and cloning terms, explained plainly
Short, honest definitions for the words you meet when you record, clean up, and clone a voice. Every term links out to the ones next to it.
Browse
Cloning and synthesis
6 termsHow a voice is captured, modelled and turned back into speech.
- Voice cloningBuilding a reusable model of one person's voice, then reading new text in it.
- Zero-shot voice cloningCloning from a few seconds of audio, with no training pass of your own.
- Instant vs professional cloneThe trade between a fast draft voice and a controlled, publishable one.
- Text-to-speech (TTS)Turning written text into spoken audio with a synthetic voice.
- Speech synthesisThe general craft of generating speech rather than recording it.
- VoiceprintThe compact numeric fingerprint that identifies one speaker to a model.
Quality and measurement
5 termsHow a generated voice is scored, and what the numbers mean.
- ProsodyRhythm, stress and intonation: the performance layer of a voice.
- Speaker similarityHow close a generated voice measures to the real speaker.
- Naturalness (MOS)A listener score for how human a sample sounds, averaged across raters.
- Word error rateHow often the words come out wrong, counted against the script.
- ArtifactA sound the model added that was never in the voice.
Audio production
6 termsThe recording terms that decide what a clone has to work with.
- Room toneThe constant quiet sound of a space with nobody speaking in it.
- Noise floorThe level of the quietest sound your recording chain always carries.
- De-essingTaming harsh s and sh sounds without dulling the whole voice.
- PlosiveThe burst of air on p and b sounds that thumps a microphone.
- RMS and LUFSTwo ways to measure loudness: raw energy, and loudness as people hear it.
- Dry recordingA take with no reverb or effects, which is what a clone wants.
Noise removal
3 termsWhat cleanup does to a recording, and where it goes wrong.
Technology
2 termsWhere the model runs, and how other tools drive it.