Final Round AI powers 25-min AI interviews with Fish Audio

Final Round AI is an AI-powered career platform that helps candidates prepare for interviews. For its mock interviews, Final Round wanted to create an experience that could sustain a real interview from start to finish. Fish Audio provides the voice layer, giving the team control over how its interviewer sounds while reducing voice and avatar costs by roughly 10x.

Final Round AI gives candidates a place to practice interviews with an AI interviewer before they face the real thing. The experience is designed to simulate a first-round interview, with candidates answering questions and responding to an interviewer over an extended conversation.

The average session lasts about 25 minutes. That makes voice a fundamental part of the product, not simply an interface. The interviewer has to maintain quality, responsiveness, and expression throughout the entire conversation.

A voice that works for a short interaction is not necessarily built for a 25-minute conversation. The interviewer needs to respond quickly, maintain consistent quality, and adapt its expression throughout the session. Small drops in quality or responsiveness become much more noticeable over a sustained interaction.

The voice also needs to work economically at the volume Final Round expects. The team wanted the flexibility to shape the interview experience around its own product, rather than accept the constraints of a packaged voice and avatar platform.

The team evaluated packaged avatar platforms that combined the avatar, voice, and infrastructure in one product. They offered a fast path to an AI interviewer, but limited how much Final Round could control the experience.

Instead, Final Round built its own interviewer avatar and integrated Fish Audio as the voice layer underneath it. The team could design the avatar and interaction around its product while Fish Audio handled the underlying voice generation.

This gave Final Round the flexibility to shape the interviewer experience without having to build and maintain its own voice infrastructure.

Separating the voice layer from the avatar experience gave Final Round more control, but it also changed the economics of running each interview.

By replacing its previous provider stack with its own avatar experience and Fish Audio for TTS, Final Round reduced voice and avatar infrastructure costs by roughly 10x. For a product built around 25-minute interview sessions, that creates meaningful room to scale usage.

The lower cost lets Final Round support longer and more frequent conversations without voice infrastructure becoming a constraint on the product.

The impact showed up in engagement. Final Round saw more than 25% longer engagement on avatar-based mock interviews compared with its earlier voice-only experience.

For an interview practice product, longer sessions mean more time for candidates to work through a complete interview and practice responding under pressure. The avatar experience gave candidates a more engaging way to use the product.

Final Round now owns the interviewer experience around its product, while Fish Audio provides the voice layer underneath it.

That gives the team the freedom to shape how its AI sounds and evolve the interviewer as the product develops, without being tied to a packaged avatar or voice experience.