Transcription serviceHindi
Hindi Transcription Services
Hindi transcription where English inside the sentence is expected rather than treated as noise, and the script policy for mixed speech is agreed before the first file.
Free sample: Send a short Hindi 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. Indian support audio is usually 8 kHz telephony, scoped as its own condition
- Script
- Devanagari throughout, or Devanagari with English spans kept in Latin, chosen before the first file
- Code-switching
- English spans transcribed as spoken, with an agreed marking convention rather than removal
- 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
Khariboli / Delhi Hindi, Bombay Hindi, Dakhini.
Who does the work
- Contributor floor
- 1,000+per market
- Markets covered
- India
- Dialects covered
- 3dialects
Coverage is drawn from India, where Spirelight already publishes a delivered Hindi dataset, so it is demonstrated rather than asserted.
Code-switching is the language, not an exception
Ordinary Hindi speech in India carries English inside the sentence: nouns, whole clauses, discourse markers, entire turns in a support call. Treating those spans as contamination to be cleaned produces a transcript of a language nobody speaks, and a model trained on it breaks on the first real call it hears.
The decision that matters is script. English spans can be written in Devanagari as pronounced, kept in Latin as written, or tagged so a downstream step can choose. All three are defensible and all three produce different training data, so Spirelight settles it with you at scoping and holds it across every batch instead of leaving it to the annotator in front of the file.
Hindi and Urdu split at the register, not at the grammar
Spoken Hindi and spoken Urdu are close to one language: the same grammar, the same rhythm in an ordinary conversation. The split shows up at the literary register and the script. Formal Hindi draws its vocabulary from Sanskrit, formal Urdu from Persian and Arabic, and Dakhini and Bombay Hindi carry Perso-Arabic loanwords into everyday speech rather than reserving them for the formal end.
That is a scoping decision, not an annotator's habit. Written as spoken, a borrowed word stays borrowed; corrected toward the Sanskritised standard, it reads more formal than the recording is. An annotator trained only on broadcast Hindi makes that correction by instinct, the way an unfamiliar transcriber cleans English out of a support call. Spirelight fixes the register convention in the style guide and staffs by the variety in the audio.
Numbers, names and the variety in the recording
Indian audio counts in lakh and crore, reads phone numbers digit by digit in patterns a Western convention does not cover, and mixes date formats inside one conversation. Proper nouns are the other recurring disagreement, because a name heard perfectly clearly still has more than one accepted Devanagari spelling.
The variety in the recording changes more than the accent. Khariboli, the Delhi-area speech the written standard is modeled on, is the default assumption behind most style guides. Bombay Hindi carries Marathi and English influence into its vocabulary, and Dakhini, spoken across Hyderabad and the wider Deccan, leans closer to Urdu than northern Hindi does. An annotator staffed to Khariboli alone will mishear or silently correct the other two, and romanized chat or voice-note text carries the same spelling ambiguity in reverse.
This service is not the right fit when
- Urdu audio in Nastaliq script. It is a separate scope with separate annotators.
- Bengali, Tamil, Telugu or Marathi audio folded into a Hindi project. Each is quoted on its own.
- Machine-only transcription with no human pass, and Hindi to English translation.
Related
Need Hindi 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.