Unverified50% confidenceFactExact time
Wake word detection typically runs lightweight models of only a few hundred KB to a few MB to continuously monitor audio streams; OpenWakeWord and Porcupine are mature open-source implementation solutions.
1
Sources
50%
Confidence
Long-term
Relevance
8/11/2026
First Seen
Sources
Related Claims
Unverified传统语音助手依赖固定唤醒词的关键词检测模型（KWS），参数量通常在100KB到1MB量级，在毫瓦级功耗下持续运行72% similarUnverifiedSilero VAD is an open-source voice activity detection solution that has gained wide popularity among developers in recent years; its model size is only approximately 2MB and it can achieve millisecond-level real-time detection without relying on a GPU.67% similarUnverifiedOpenAI开源的Whisper模型大幅提升了识别准确率，whisper.cpp、faster-whisper等移植版本将本地推理延迟压缩至实时以下66% similarUnverifiedByteDance has released Seed Live Interpret 2.0, an end-to-end simultaneous interpretation model that approaches human-level latency and accuracy for Chinese-English simultaneous interpretation and supports zero-shot voice cloning.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/731092API
curl https://kongchang.com/api/v1/knowledge/claims/731092MCP
get_claim(id=731092)