Unverified50% confidenceBenchmarkExact time

在 M1 芯片上运行 Whisper base 模型的转录速度可达音频时长的 10-15 倍，即 1 分钟音频仅需 4-6 秒完成转录

1

Sources

50%

Confidence

Long-term

Relevance

8/9/2026

First Seen

Sources

VoiceDumps: A Fully Offline, Free, and Open-Source Voice Input Tool for macOS

Related Claims

UnverifiedWhisper small模型在现代笔记本电脑的CPU上处理一段10分钟音频通常只需1-3分钟79% similar Unverified原版 Whisper 的单次处理窗口仅为 30 秒，需要通过滑动窗口拼接处理长音频77% similar UnverifiedWhisper的语音识别模型只能处理16kHz采样率的单声道PCM浮点数据格式的原始音频波形73% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/717351

API

curl https://kongchang.com/api/v1/knowledge/claims/717351

MCP

get_claim(id=717351)