The archive · AI & Models · Strategic decision · 2019–2026
Resemble AI bets MIT open-source voice cloning becomes the default TTS
Voice startup releases Chatterbox under MIT, hits GitHub trending #1 and 26.2k stars; sells hosting, watermarking and detection.
Resemble AI
What the business is
Resemble AI is a generative voice company that clones and synthesizes voices with consent controls, PerTh audio watermarking and its Resemble Detect authenticity model, and since 2025 publishes the open-source Chatterbox text-to-speech family.
Starting capital：$8M Series A led by Javelin Venture Partners with Craft and Ubiquity, announced 2023-07-12, bringing total raised to $12M at the time (TechCrunch).
How it started
Resemble AI was founded in 2019 by Zohaib Ahmed, a former Magic Leap lead software engineer, and Saqib Muhammad, after they noticed video-game voice lines could not keep up with frequent game updates. The company required consent clips before cloning a voice, enforced strict usage guidelines, and built PerTh watermarking plus Resemble Detect early, positioning itself as the responsible player in a market it acknowledged was controversial.
What happened
In April 2025 Resemble shipped Chatterbox, its first production-grade open-source TTS model, under an MIT license. The repo grew 10.4x in Q3 2025 per the ROSS Index, and star-history shows it reached GitHub trending #1 on December 17, 2025, with repeated top-10 trending days into March 2026. On December 26, 2025 the company released Chatterbox Turbo: first audio in under 150 milliseconds, zero-shot cloning from five seconds of audio, built-in PerTh watermarking for verifying AI-generated speech, MIT-licensed, and available on Hugging Face, RunPod, Modal, Replicate and Fal.
How it ended up
Live and scaling its bet: as of September 2026 the MIT-licensed Chatterbox family spans multilingual and Turbo models, the main repo holds 26.2k stars, and Resemble continues to sell hosted voice plus AI-generated-content detection to enterprises and regulated industries.
Background
Resemble AI is a generative voice company founded in 2019 by Zohaib Ahmed and Saqib Muhammad, who started from a simple observation: video-game voice lines could not keep up with frequent game updates. The company grew from game voice-overs into dubbing, personalized voice messages and real-time conversational agents, and it differentiated early on consent and safety — requiring a consent clip before cloning and building PerTh watermarking and Resemble Detect, an authenticity model that scores whether audio is real or manipulated.
In 2025 Resemble reversed the usual voice-AI playbook. It released Chatterbox, its production-grade text-to-speech model, under the permissive MIT license, then followed in December 2025 with Chatterbox Turbo — first audio in under 150 milliseconds, zero-shot voice cloning from five seconds of material, built-in PerTh watermarking, and free commercial use. The Decoder and PingWest both covered the release as a direct challenge to commercial voice platforms like ElevenLabs, aimed at developers building real-time agents, customer support, games, avatars and social products.
The open-source bet produced the traffic a closed API could not. The resemble-ai/chatterbox repo grew 10.4x in Q3 2025 per Runa's ROSS Index, reached GitHub trending #1 on December 17, 2025, and held 26.2k stars by September 1, 2026, with 21 total trending days. The models spread across Hugging Face, RunPod, Modal, Replicate and Fal, so developers could test and ship them without touching Resemble's paid platform.
Resemble's strategy is to let open source capture the developers and monetize the enterprises: a hosted service with low-latency deployment, watermarking that works on its own generated speech, and detection tools for regulated industries worried about deepfakes. As of September 2026 the company was still live, with the open Chatterbox family spanning multilingual and Turbo variants and a commercial security business built around AI-generated content.
What has to be true
- Open source flipped the trust question: showing the model code under MIT gave developers proof of quality and control, converting skeptics of a voice-cloning vendor into adopters.
- The release was timed to the real-time agent wave: under-150ms first audio and five-second cloning directly targeted voice agents, support bots, games and avatars — not just offline dubbing.
- Safety features were bundled into the giveaway: built-in PerTh watermarking made the open model usable by regulated industries that would otherwise reject unmarked synthetic speech.
- Distribution multiplied reach: launching on Hugging Face, RunPod, Modal, Replicate and Fal put the model in front of developers who never visit the vendor's site.
What can be applied
Giving away the crown jewels can expand the market you sell into: an MIT-licensed, best-in-class model built developer trust and mindshare that a closed API alone never earned.
Aftermath
As of September 2, 2026, Resemble AI is live with an open-plus-commercial voice strategy: the MIT-licensed Chatterbox family includes multilingual and Turbo models, the GitHub repo holds 26.2k stars with a #1 trending day on December 17, 2025, and models are served via Hugging Face, RunPod, Modal, Replicate and Fal. It still sells hosted voice plus PerTh watermarking and Resemble Detect to enterprises and regulated industries. The open question is whether giving away its best model cannibalizes hosted TTS revenue faster than trust tooling and enterprise support replace it.
Sources
- Chatterbox Turbo: Kostenloses Audio-Modell klont Stimmen in wenigen Sekunden
- Resemble AI 推出开源语音合成模型 Chatterbox Turbo
- resemble-ai/chatterbox star history
- Voice cloning platform Resemble AI lands $8M
- Top trending open-source startups in Q3 2025 (ROSS Index)
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card