# silence threshold endpointing

Published articles for silence threshold endpointing.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Semantic VAD: turn detection that uses meaning, not silence

DevFeed: [Semantic VAD: turn detection that uses meaning, not silence](<https://devfeed.tech/articles/semantic-vad-turn-detection-that-uses-meaning-not-silence-81243.md>)

Original publisher: [Read original article](<https://gradium.ai/blog/semantic-vad>)

Author: Gradium

Published: 2026-06-02T00:00:00Z

Content type: tutorial

Language: en

Sources: [Gradium](<https://devfeed.tech/sources/gradium.md>)

Topics: [speech-to-speech](<https://devfeed.tech/topics/speech-to-speech.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [acoustic-vad-vs-semantic-vad](<https://devfeed.tech/tags/acoustic-vad-vs-semantic-vad.md>), [api](<https://devfeed.tech/tags/api.md>), [assemblyai-semantic-endpointing](<https://devfeed.tech/tags/assemblyai-semantic-endpointing.md>), [barge-in-detection](<https://devfeed.tech/tags/barge-in-detection.md>), [delay-in-frames](<https://devfeed.tech/tags/delay-in-frames.md>), [end-of-turn-detection](<https://devfeed.tech/tags/end-of-turn-detection.md>), [end-of-utterance-detection](<https://devfeed.tech/tags/end-of-utterance-detection.md>), [endpointing](<https://devfeed.tech/tags/endpointing.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [eot-detection](<https://devfeed.tech/tags/eot-detection.md>), [eou-detection](<https://devfeed.tech/tags/eou-detection.md>), [gradium-stt](<https://devfeed.tech/tags/gradium-stt.md>), [inactivity-probability](<https://devfeed.tech/tags/inactivity-probability.md>), [openai-semantic-vad](<https://devfeed.tech/tags/openai-semantic-vad.md>), [real-time-stt-turn-detection](<https://devfeed.tech/tags/real-time-stt-turn-detection.md>), [semantic-endpointing](<https://devfeed.tech/tags/semantic-endpointing.md>), [semantic-vad](<https://devfeed.tech/tags/semantic-vad.md>), [semantic-voice-activity-detection](<https://devfeed.tech/tags/semantic-voice-activity-detection.md>), [send-flush-stt](<https://devfeed.tech/tags/send-flush-stt.md>), [silence-threshold-endpointing](<https://devfeed.tech/tags/silence-threshold-endpointing.md>), [silero-vad](<https://devfeed.tech/tags/silero-vad.md>), [telephony-vad](<https://devfeed.tech/tags/telephony-vad.md>), [turn-detection-voice-agent](<https://devfeed.tech/tags/turn-detection-voice-agent.md>), [turn-taking-voice-agent](<https://devfeed.tech/tags/turn-taking-voice-agent.md>), [voice](<https://devfeed.tech/tags/voice.md>), [voice-agent-interruption](<https://devfeed.tech/tags/voice-agent-interruption.md>), [voice-agent-latency](<https://devfeed.tech/tags/voice-agent-latency.md>), [webrtc-vad](<https://devfeed.tech/tags/webrtc-vad.md>)

### AI overview

Semantic voice activity detection (VAD) uses linguistic context to estimate whether a speaker has finished a turn, addressing interruptions caused by acoustic detectors that treat pauses as endpoints. The article explains Gradium STT's multi-horizon inactivity predictions, tuning delay and thresholds, and using send_flush() to emit pending transcript text at turn boundaries.

### Source excerpt

Acoustic VAD answers "is there a voice right now?" Semantic VAD answers "is the user done talking?" Here's why the distinction decides whether a voice agent cuts users off, how Gradium STT emits multi-horizon turn-completion predictions every 80 ms, and how to tune delay_in_frames, horizon, and flushing for your use case.