Developer Doug created 'erm', a command-line tool that automatically removes speech disfluencies (ums, uhs, ers) from audio files using Whisper's speech recognition and custom audio processing. The tool addresses key challenges like Whisper's inconsistent filler word detection and audio artifacts from splicing, providing a local solution for content creators. While innovative for its niche, it's a specialized utility rather than a major technical breakthrough.
Background
Speech disfluencies like 'um' and 'uh' are common in spoken language but often need to be removed for professional audio content. Traditional manual editing is time-consuming, creating demand for automated solutions.
- Source
- Lobsters
- Published
- Jun 13, 2026 at 01:23 AM
- Score
- 5.0 / 10