Gradium

Gradium is an AI voice startup based in Paris. It builds ultra-low latency voice AI models and real-time, scalable voice APIs for developers, covering expressive text-to-speech, accurate speech-to-text and high-fidelity voice clones. Its technical work spans generative audio, neural audio codecs, audio language models and real-time inference.

The company's product line has two parts. The Voice AI API platform delivers real-time voice services to developers with ultra-low latency as its stated key metric. Alongside it, Gradium has developed foundational audio language models: an audio-native family that unifies generation, transcription, transformation and dialogue within a single neural architecture, aiming to produce realistic, emotionally expressive speech efficiently and at scale.

The founding team collectively invented and published the methods and algorithms behind many of the voice and audio models in use today. Gradium reports serving customers in production within months of launch, and targets verticals including healthcare, customer support, market research, gaming, digital advertising and developer tools.

Technical domains include text-to-speech, speech-to-text, voice cloning, speech synthesis and audio language models. The company states a focus on making high-quality voice AI affordable and broadly available, with developers as its primary users.