What we can take on
Verbatim transcription
Every false start, filler and repetition, for ASR training.
Clean read
Readable transcripts for search, analytics and subtitles.
Speaker diarisation
Who spoke when, with consistent speaker identity across a session.
Timestamping
Word-level or segment-level alignment in CTM or JSON.
Phonetic transcription
IPA transcription for text-to-speech and linguistic research.
Domain transcription
Clinical, legal and financial audio handled by subject-matter experts.
The types of data transcription we run
The full breakdown, not the headline categories. Anything not on this list is a scoping call away.
Transcription style. Verbatim, every disfluency kept, Intelligent verbatim, Clean read, Phonetic transcription (IPA).
Structure. Speaker diarisation, Word-level timestamping, Segment-level timestamping, Overlap and cross-talk marking.
Domain. Clinical dictation, Legal proceedings, Financial and earnings calls, Broadcast and media.
Output. CTM alignment files, SRT / VTT subtitles, Structured JSON, Plain text against a style guide.
The forms the data takes
What you actually receive: file types, structure and the metadata that travels with it.
- Verbatim transcripts
- Every word as spoken, including fillers, false starts and non-speech tags.
- Clean transcripts
- Readable text with disfluencies removed, per your style guide.
- Timestamps
- Word-level or segment-level time codes in JSON, SRT, VTT or TextGrid.
- Speaker-labelled
- Turn-by-turn speaker IDs with diarisation, for calls and meetings.
- Code-switched
- Mixed-language speech transcribed as spoken, with language tags per span.
- Validation reports
- Audio-quality flags and transcript-to-audio agreement scores.
How the data is collected
From the first brief to the batch you receive, step by step.
- 01
Audio from you or from our collection
We transcribe audio you supply, or audio we recorded for you, and keep each transcript tied to its source file.
- 02
Machine pre-fill, human correction
A first-pass transcript is generated and then corrected by native-speaker transcribers working in a waveform-linked editor.
- 03
Style guide per project
Verbatim or clean, how to mark fillers, laughter, overlap, numbers and code-switching: written down and agreed before the first batch.
- 04
Speakers and time
Speaker labels and word or segment timestamps are added where the project calls for them, with diarisation checked by a second person.
- 05
Reviewed against a word-error gate
A second transcriber reviews a sample and an automated word-error check blocks batches that miss the agreed bar.
How a project runs
- 01
Scoping call
We agree volume, languages, demographic quotas, formats and what counts as a rejected item.
- 02
Guidelines and gold set
Written guidelines with worked examples, plus a gold set every contributor is tested against.
- 03
Pilot batch
A small paid batch you accept before anything scales. Most guideline problems surface here.
- 04
Production
Work flows through capture checks, automated QC, peer review and expert review.
- 05
Delivery
Incremental batches in your format with a data card, QA report and agreement scores.
- 06
Iteration
Rejections feed back into the guidelines so the same mistake is not made twice.
How quality is checked
Checks that run during the work, not a sample at the end.
- Native-speaker transcribers for each language and dialect
- Second-pass review on a risk-weighted sample
- Automated word-error check before delivery
- Redaction and PII verification where calls contain personal data
- Style-guide adherence scored per transcriber
Industries that use it
Where teams bring us data transcription, and what they use it for.
- TelecomSupport-call transcription in many languages.
- Banking and financeCall-centre and advisory call transcripts, with redaction.
- HealthcareDictation and consultation transcripts under a privacy protocol.
- AutomotiveIn-car command and conversation transcripts.
- BPO and call centresQuality-assurance transcripts at volume.
- Media and entertainmentSubtitle, caption and dubbing source text.
Why teams pick us for data transcription
Native dialect routing
Bengali from Murshidabad does not go to a Dhaka transcriber by default.
Automatic WER gate
Machine check before human review, so reviewers spend time on real problems.
Style guides honoured
Your conventions for numbers, names and disfluencies, applied consistently.
Secure handling
Clinical and financial audio stays inside controls your risk team can sign off.
