Case Study: Resemble AI benchmarks Chatterbox with Podonos

Resemble AI benchmarks Chatterbox in under 12 hours with Podonos

Resemble AI, a company specializing in AI voice technology, faced the challenge of objectively benchmarking its new text-to-speech model, Chatterbox, against a leading competitor like Eleven Labs before its open-source release. They needed to move beyond subjective claims and provide transparent, data-driven proof of the model's performance to establish trust within the AI community. To achieve this, they used the evaluation service from Podonos.

Podonos provided a fast and automated evaluation solution, enabling Resemble AI to conduct controlled A/B tests against Eleven Labs using real-world prompts and receive detailed feedback from diverse evaluators in under 12 hours. The results confirmed Chatterbox's competitive strengths, particularly in naturalness. This data-backed confidence allowed Resemble AI to successfully open-source the model on Github and Hugging Face, enhancing its credibility upon release.

Resemble AI