How we chose the voices of Coda

Coda's recent launch involved not only the challenge of training the model but also the critical task of selecting and categorizing the top voices from over 180 candidates to ensure alignment with real-world use cases. Partnering with Podonos, an audio evaluation platform, Coda conducted over eight thousand pairwise listener judgments to refine its voice catalog, classifying voices into four styles: professional, formal, casual, and energetic. These categories were crafted in response to customer feedback and are crucial for large enterprises where incorrect voice selection can have significant financial implications. The selection process involved a rigorous methodology, utilizing pairwise comparisons rather than ratings, and employed Bradley-Terry scoring to achieve a statistically robust ranking. Podonos was chosen for its ability to provide precise evaluations with a vetted panel, contrasting with platforms like Rapidata, which offer broader but less controlled assessments. The voices identified as winners in each category reflect versatility, capable of transitioning across different registers to meet varied customer needs.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.