Speech Assist · preview
Clearer conversations. Same voice.
Softens accents bidirectionally in real time while preserving each speaker's natural voice and authenticity, clarity for customers and agents without cloning anyone.
Bidirectional, agent side and customer sidePreserves each speaker's own voice, no cloning, no enrollment~200 ms target latency, CPU-only
~200 msTarget algorithmic latency, CPU-only
BidirectionalAgent-side and customer-side conversion
PreviewVoice-conversion model in development
What it does
Real-time accent softening for agents.
Agent-side conversion
Softens the agent's accent in real time while keeping their own voice, no cloning, no enrollment.
Customer-side conversion
Neutralizes the inbound customer accent the agent hears, reducing cognitive load and repeat requests.
Voice preserved
Timbre, tone, and emotion are kept intact. Only pronunciation is reshaped.
How it runs
Three steps to live.
1
Attach
Runs as an in-path add-on on the same audio stream as noise cancellation.
2
Select target
Choose the target accent per queue or per agent.
3
Run
Clearer conversations with no workflow changes.
See Accent Conversion on a real call.
A live demo on a real phone line, in Arabic and English.