Gradium
Updated: December 13, 2025
The team
Country: France | Funding: $70M (+)
Founded: 2025
Website: https://gradium.ai
Gradium creates audio-language AI models, designed to generate and recognize voices with ultra-low latency. The startup's goal is to create a voice API for developers that works in real time. Its voice cloning feature allows to create custom AI voices, including your own (you simply need to upload a 10-second audio sample for speech synthesis). The models have multilingual support: English, French, German, Spanish and more languages to be added in the future. Gradium competes with leading LLMs (OpenAI, Anthropic, Meta Llama, and Mistral) which are providing voice and speech models. However, Gradium cites that its advantages are hyper-realistic voice reproduction and accuracy.
Founded: 2025
Website: https://gradium.ai
Gradium creates audio-language AI models, designed to generate and recognize voices with ultra-low latency. The startup's goal is to create a voice API for developers that works in real time. Its voice cloning feature allows to create custom AI voices, including your own (you simply need to upload a 10-second audio sample for speech synthesis). The models have multilingual support: English, French, German, Spanish and more languages to be added in the future. Gradium competes with leading LLMs (OpenAI, Anthropic, Meta Llama, and Mistral) which are providing voice and speech models. However, Gradium cites that its advantages are hyper-realistic voice reproduction and accuracy.
