Camb AI, a Dubai-based startup, has released Mars5, a powerful AI model for voice cloning that claims to offer a higher level of realism than its competitors. Mars5 can emulate voices with prosody, capturing complex parameters such as rhythm, emotion, and intonation. The model supports over 140 languages, including low-resource ones like Icelandic and Swahili. What sets Mars5 apart is its ability to capture diverse emotional tones and pitches, making it suitable for content that has been traditionally difficult to convert into speech, such as sports commentary, movies, and anime. The company’s co-founder and CTO, Akshat Prakash, stated that Mars5’s level of prosody and realism is unprecedented, and the model is capable of capturing nuances that other models cannot.
In a unified platform, Mars5 combines voice cloning and text-to-speech conversion capabilities, allowing users to upload an audio file and provide text content, and the model will synthesize the provided text as speech using the speaker’s voice as a reference. The company claims that Mars5 outperforms popular open and closed-source speech synthesis models, including those from Metavoice and ElevenLabs.
Camb AI is also planning to release another model called Boli, which is designed to enable translation with contextual understanding, correct grammar, and apt colloquialism. Both Mars5 and Boli work with 140 languages on the company’s proprietary platform Camb Studio, and the company is providing these capabilities as APIs to enterprises, SMEs, and developers.
