In the catalog since Aug 19, 2026
Resemble AI
0index_Resemble AICompany
Voice studio from Resemble AI: clones a voice from a short sample, narrates text, and immediately checks someone else's recording for forgery. The bet is on the open Chatterbox model and a watermark in every file.
Facts
- The open Chatterbox model is released under the MIT license: voice cloning without fine-tuning from a 5-20 second sample, 23+ languages, and about 200 ms latency (resemble.ai/learn/models/chatterbox, Aug 19, 2026).
- The Chatterbox repository opened on Apr 23, 2025, with the first v0.1.2 release on Jun 13, 2025; as of Aug 19, 2026, it had 26,046 GitHub stars (github.com/resemble-ai/chatterbox).
- The DETECT-World detector family identifies synthetic content in audio, video, and images; the product page claims 99.5% accuracy for audio, 98% for video, and 96% for images (resemble.ai, Aug 19, 2026).
- Every generation is labeled with a PerTh Multimodal watermark, with claimed C2PA compatibility; call protection and a Chrome extension are sold separately (resemble.ai, Aug 19, 2026).
The full card is in the works: the "worth it or not" verdict, pricing and a comparison with neighbors will appear as it moves through the card factory.