Curiosity · AI Model
Resemble Rapid Voice Cloning
Rapid Voice Cloning is Resemble AI's fast voice-cloning technology, generally available from 2023 and updated through 2024-25. It creates a custom TTS voice from as little as 10 seconds of clean reference audio, supporting 149 languages, emotion controls, and real-time streaming inference. Resemble pairs cloning with PerTh Watermarker — a neural watermark embedded in every generated clip — and a deepfake-detection API (Resemble Detect) to mitigate misuse. Targeted at customer-support, gaming, and content production at enterprise scale.
Model specs
- Vendor
- Resemble AI
- Family
- Resemble
- Released
- 2023-06
- Context window
- 4,096 tokens
- Modalities
- text, audio
- Input price
- n/a
- Output price
- n/a
- Pricing as of
- 2026-04-20
Strengths
- Very few-shot cloning from short clips
- Wide language coverage (149)
- PerTh watermarking on every output for provenance
- Dedicated deepfake-detection API (Resemble Detect)
Limitations
- Closed weights — API only
- Expressive quality depends heavily on reference audio cleanliness
- Enterprise-priced — less attractive for solo creators
- Ethical/legal review required before cloning third parties
Use cases
- IVR and customer-support personalisation
- Gaming NPC voice assets
- Multi-language dubbing of marketing content
- Accessibility tools with personalised voices
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| Minimum reference audio | ≈10 seconds | 2024-01 |
| Language support | 149 languages | 2024-12 |
Frequently asked questions
What is Resemble Rapid Voice Cloning?
A Resemble AI service that clones a voice from ~10 seconds of reference audio and delivers a streaming TTS endpoint across 149 languages.
How does Resemble mitigate misuse?
Every generated clip is watermarked with PerTh, and Resemble ships Resemble Detect — a deepfake-detection API — to identify synthetic speech.
Is Rapid Voice Cloning legal?
Legality depends on jurisdiction and consent. Resemble's terms require the user to have rights to the source voice; platform-level checks help enforce this.
Sources
- Resemble AI — Rapid Voice Cloning — accessed 2026-04-20
- Resemble AI — PerTh Watermarker — accessed 2026-04-20