Data labeling
Audio transcription and speech data annotation for ASR
Verbatim and clean transcripts, word and segment timestamps, speaker diarization and sound events, by native speakers who hear accents and code-switching as natural speech, not noise.
- Pilot phase
- Two or more review layers
- You own every label
Real people don’t speak textbook language.
Speech models trained on clean, single-language audio struggle with real callers: accents, background noise, fillers and sentences that switch between Swahili and English halfway through. The fix is transcription that captures how people actually talk.
Our transcribers follow written conventions for language switches, overlapping speech, fillers and non-speech events, so every person makes the same call on the same audio. Sensitive details such as ID numbers and payment codes can be redacted before delivery.
Used for
- Speech recognition (ASR) training and evaluation
- Call-center analytics
- Voice assistants
- Podcast and media transcription
- Field recordings and research interviews
Industries: Generative AI & LLMs · Finance & insurance
What our audio transcription covers
Combine any of these in one workflow and one guideline.
Verbatim & clean transcription
Every word and filler, or clean read text, to your style guide.
Timestamps
Word-level or segment-level timing for alignment and subtitles.
Speaker diarization
Who spoke when, including overlaps and interruptions.
Language & code-switch tags
Language labels for English, Swahili and Sheng segments.
Sound-event tagging
Noise, music, laughter and other non-speech events.
PII redaction
Removal of ID numbers, phone numbers and payment codes.
Formats and tools
Delivery formats
- JSON with timestamps
- SRT / VTT
- TextGrid
- CSV
- Custom schema
Tools we work in
- Label Studio
- ELAN
- Your in-house tool
Pricing
Audio transcription from $0.80 per audio minute, and from $0.60 at volume, including timestamps.
How we keep it accurate
Every project runs on the same four-check QA process. See how our QA works.
Word error rate audits
Reviewers re-transcribe samples so WER is measured on every batch.
Convention checks
Consistent handling of fillers, overlaps and code-switching across the whole team.
Redaction pass
A separate pass for personal data before any file leaves our hands.
How a project runs
Scope and guidelines
We review a sample, agree classes, edge cases and acceptance criteria, and sign your NDA.
Pilot phase
Pilot phase helps you judge quality, speed and communication.
Certified team
Annotators pass a gold-set test before they are cleared for production.
Production and reporting
Batches with QA reports, in your format, on an agreed schedule.
Audio transcription questions
Which languages and accents do you transcribe?
English, Swahili and Sheng, including code-switched speech and a wide range of East African accents. Other languages are assessed case by case.
How do you price audio transcription?
Per audio minute. Standard transcription with timestamps starts at $0.80 per minute. See the rate card.
Can you handle noisy recordings?
Yes. Inaudible segments are marked using your conventions rather than guessed, so your model doesn’t learn invented words.
Do you redact personal data?
Yes. We can run a dedicated redaction pass for ID numbers, phone numbers, payment codes and names.
Send a sample. See the labeling quality first.
Share a few files and a line about your goals. A project lead replies within one business day with an approach, a timeline and a price.