Body-Cam Tagging & Transcription

I’m creating high-quality training data for a speech-recognition model that relies on police body-worn camera audio. The work is split into two tightly linked tasks and I need someone who is already comfortable with both speaker identification and word-for-word transcription.

Task flow
Job 1 – Speaker roster
• Listen through each full recording once, numbering every new voice in the exact order it first appears.
• Give each speaker a concise role label plus a brief description.
• Clip three short (2-8 s) reference samples for every speaker.
• Flag any two voices that sound alike so the model can learn the difference.

Job 2 – Clip-level transcript
• I supply pre-cut, pre-filled segments; you correct timing and text so every audible word is captured.
• Keep slang, fillers and profanity exactly as spoken.
• Mark overlap, background radio traffic, or unintelligible portions with the required tags (( )).
• No guessing—if you don’t hear it, tag it.

What you should bring
• Proven experience handling both speaker identification and transcription.
• Familiarity with police, emergency, interview, meeting, podcast, or other real-world recordings.
• A preference for mostly clear audio is fine; if a clip is too distressing you can skip it with no penalty.
• Good closed-back headphones and the focus to deliver error-free work; submissions with multiple mistakes won’t be paid and may be rejected.

Deliverables
1. Completed speaker roster (.xlsx or .csv) with numbered voices, roles, descriptions, match flags, and reference clip filenames.
2. Corrected transcript files for all assigned clips, following the tagging guidelines exactly.

If the scope and detail above match your skill set, let’s move forward and get the data model-ready.