Transcription serviceKorean
Korean Transcription Services
Korean transcription that keeps the speech level the speaker actually used, covers Gyeongsang and Jeolla audio, and holds one spacing convention across every batch.
Free sample: Send a short Korean audio clip and we return a transcript in the format and orthography you specify, at no cost, before you commit to a project.
Specification
- Audio
- Wideband, narrowband, mono, or separated channels
- Register
- Speech levels and honorific forms preserved as produced, or normalized to a stated target
- Spacing
- One spacing convention fixed in the style guide and scored against it
- Accuracy
- Target agreed per project and measured on a held-out slice of your own recordings
- Review
- Second-pass review on a sampled percentage, rate agreed up front
Dialects and varieties covered
Seoul, Gyeongsang, Jeolla.
Who does the work
- Contributor floor
- 1,000+per market
- Markets covered
- South Korea
- Dialects covered
- 3dialects
Coverage is drawn from South Korea, where Spirelight already publishes a delivered Korean dataset, so it is demonstrated rather than asserted.
Speech level is content, not politeness decoration
Korean marks the relationship between speakers in the verb ending. A customer who slides from the polite ending into plain form partway through a call has done something a model should be able to detect, and a transcriber who tidies every ending into one formal register has erased it. Honorific verb stems and address terms carry the same signal.
Verbatim delivery at Spirelight preserves the ending as produced, including the shifts inside a single turn. If you want a clean read instead, the scope states which endings are normalized and to what, so it is a documented choice rather than an annotator's habit applied unevenly across a batch.
Spacing is the quiet source of disagreement
Hangul is written with word spacing whose rules permit more than one defensible answer in ordinary cases, particularly around dependent nouns and compound verbs. Two competent transcribers will space the same utterance differently, and if you score by token that difference lands in your error rate without anyone having misheard a word.
Spirelight fixes a spacing convention up front and measures against it. The same rule covers numerals: whether a figure is written in digits or spelled out, and whether the native or Sino-Korean series is written as heard. Jeju audio is scoped on its own, because it is far enough from mainland Korean to need its own speakers.
This service is not the right fit when
- North Korean speech. Spirelight does not staff it, and saying so is more useful than approximating it.
- Korean to English translation or timed subtitle delivery.
- Machine-only transcription with no human pass.
Related
Need Korean transcription scoped to your project? Send the audio conditions, volume, and turnaround you need. We will reply with a scoped quote and, if it helps, a free sample.