Together AI Voice Finder Simplifies Voice Selection

Finding the perfect voice for an AI application just got significantly easier. Together AI has launched Voice Finder, a new utility designed to help developers quickly navigate its extensive catalog of over 600 synthetic voices.

The challenge with existing Text-to-Speech (TTS) platforms is often the manual effort required to sift through numerous voice options. Provider catalogs can be overwhelming, with documentation rarely offering clear guidance on which voice best suits specific use cases like a fintech support agent or a game show host. Voice Finder aims to solve this by offering a more intuitive and efficient search experience.

Search by Description or Sound

Developers can now search for voices using natural language prompts, describing the desired characteristics such as "a calm female voice for a meditation app." Alternatively, users can upload a short audio sample of a voice they have in mind, and Voice Finder will surface similar-sounding options.

This functionality leverages an "omni-model" that has analyzed each of the 600+ voices across more than 15 attributes. These include pitch, gender, accent, language, age, emotion, and speaking style, allowing for granular filtering.

Streamlining AI Voice Agent Development

The quality of the voice is critical for the user experience of AI voice agents. A healthcare intake agent, for instance, should not sound the same as an entertainment companion. Together AI's platform integrates Speech-to-Text (STT), Large Language Models (LLMs), and TTS, enabling end-to-end voice agent development with low latency. Voice Finder specifically targets the voice selection bottleneck within this workflow, making AI voice agent development more accessible.

This tool supports AI voice agent development, a crucial component for creating natural and engaging conversational AI experiences. Companies like ElevenLabs are also innovating in this space, as seen in ElevenLabs Gives Chat Agents a Voice.

Voice Finder is available now, with the goal of reducing the time from identifying a need to selecting a suitable voice, thereby accelerating the development cycle for AI-powered applications.