Keyterm Filtering for Voice AI

Name: Keyterm Filtering for Voice AI
Availability: InStock
Author: mayowa_osibodu

by mayowa_osibodu·May 5, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidSolve My ProblemNiche Gem

Cuts keyterm hallucinations by 60% before audio hits Deepgram.

Strengths

•Pre-filtering architecture prevents hallucinations rather than fixing them post-hoc.
•Supports raw PCM streaming for real-time AudioWorklet integration.
•Addresses a specific, painful edge case in non-English STT contexts.

Weaknesses

•Currently limited to en-IN and hi-IN, restricting global utility.
•Adds an extra network hop and latency to the transcription pipeline.

Post Description

Keyterm prompting is a valuable way to help your STT better recognize unique terms like brand names etc, but for non-English languages/non-standard accents, providers like Deepgram tend to hallucinate keyterms in STT transcripts. So the output transcript contains the given keyterms, even when those keyterms are not present in the input audio.

I'm currently collecting feedback to improve this product. Right now it cuts down keyterm hallucinations by about 60% on in-house test data, so I'm curious to see how it performs in public.

The product is free to use while in beta (Hindi and Indian-accented English are supported). Would love to hear how it performs on your data. Feel free to drop a comment if you’re interested in features like additional language support, streaming and self-hosting.