How it works
VidPickr pulls the audio-only stream of the video you paste, hands it to our speech-recognition pipeline, and produces a clean transcript with timestamps. First run takes a moment to warm up; the same machine then starts transcribing instantly on subsequent videos.
Why use this over other transcription tools?
Most transcription sites cap you to a few minutes per day, charge per minute of audio, or hand back lossy auto-captions disguised as new transcripts. VidPickr runs Whisper locally, so one Plus subscription at $1/month covers every video, every length and every supported language, with real word-level timestamps in the SRT and VTT output and no per-minute metering. The tradeoff is honest: the work happens on your CPU, so a long video takes longer than it would on a rented GPU.
Speed expectations
- Apple Silicon (M1/M2/M3/M4): ~3-5× realtime on Whisper Base. A 30-minute video transcribes in 6-10 minutes.
- Recent Intel + dedicated GPU: similar, with WebGPU. Without WebGPU, ~1-2× realtime.
- Older laptops / mobile: works but slower; consider Whisper Tiny (39 MB) for short videos under 5 min.
What about subtitles that already exist?
If the video creator uploaded subtitles or YouTube generated auto-captions for it, you don’t need this tool — just use the subtitle downloader which exports those directly. Use Transcribe when there are no subtitles, when the auto-captions are wrong, or when you want a different language via the “Translate to English” mode.