Starting a Transcription
To transcribe a video, select it in the Evidence Panel and click Run Analysis > Transcription in the toolbar. Alternatively, right-click the video and choose Transcribe.
FrameCounsel uses a locally-optimized Whisper model running on Apple Silicon's Neural Engine. No audio data ever leaves your machine.
Language Settings
By default, FrameCounsel auto-detects the spoken language. To set a specific language:
- Open Settings > Transcription.
- Select a language from the dropdown. Over 50 languages are supported.
- If the video contains multiple languages, enable Multi-language mode for automatic switching.
Tip: Forcing a specific language (rather than auto-detect) can improve accuracy by 10-15%, especially for accented English or bilingual conversations.
Transcription Speed
Transcription speed depends on your hardware and the video length:
- M4 Pro/Max (recommended): Approximately 1 minute of transcription per 15 minutes of video
- M4: Approximately 1 minute per 10 minutes of video
- M2/M3 Pro/Max: Approximately 1 minute per 6 minutes of video
- M1: Approximately 1 minute per 3 minutes of video
A progress bar shows real-time status. You can continue working on other tasks while transcription runs in the background.
Reviewing and Editing Transcripts
Once complete, the transcript appears in the Findings Panel with timestamps. Click any line to jump to that moment in the video. To edit:
- Double-click a transcript line to enter edit mode.
- Make your corrections. The original AI-generated text is preserved in the audit log.
- Press Enter to save. Edits are marked with an "edited" indicator.
Accuracy Tips
For the best transcription results:
- Use videos with clear, direct audio. Body cameras worn on the chest produce better results than dashboard cameras with ambient noise.
- If audio quality is poor, try the Enhanced Audio Processing option in the transcription settings, which applies noise reduction before transcription.
- For multi-speaker scenarios, enable Speaker Diarization to label different speakers (Officer 1, Subject, Bystander, etc.).
Warning: Transcription accuracy is not 100%. Always review the full transcript against the video before using it in court filings or reports. The transcript is a tool to aid your review, not a substitute for listening to the audio.