Anam Solves Turn-Taking for Unscripted Avatar Conversation with Flux STT

Anam, a provider of photorealistic interactive avatars, adopted Deepgram’s Flux speech-to-text system to improve real-time conversational turn-taking while retaining self-hosted deployment on its Kubernetes infrastructure. In A/B testing against home-grown, competitor, and open-source alternatives, Flux reportedly reduced unwanted avatar interruptions by about 50% without increasing p95 latency, while better handling pauses, background noise, and varied speaking styles. Anam deployed the system on H100 GPUs and rolled it out to all turnkey users after a roughly two-week self-hosted production process, with customer complaints about English-session turn-taking falling close to zero. The company says the improvements have encouraged some customers to move from their own speech-to-text setups to Anam’s managed experience, and it plans to expand use of Flux as language support grows while also considering Deepgram’s text-to-speech technology.

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.