Pricing
ASR Training Data: cost and pricing in India
Indicative rate bands for asr training data, the five specification decisions that move the number most, and how a fixed quote is put together from a written spec.
- Pricing model
- Per delivered unit, fixed against spec
- Minimum
- ~50 hours or equivalent per language
- Turnaround
- Transcription adds roughly 1-2 weeks per 100 hours after recording, per language.
- Quote time
- One working day
What drives the price of asr datasets
Transcribed speech corpora built to train and evaluate automatic speech recognition, with verbatim transcription, timestamps and per-token language tagging where code-mixing occurs.
- Language tier: widely spoken languages have deeper recruitment pools and cost less per unit than low-resource ones
- Demographic narrowness: a 50/50 gender split across three age bands is routine; a narrow cohort multiplies recruitment effort
- Audio — 16 kHz or 48 kHz PCM WAV. Moving this off the standard changes the unit cost directly.
- Transcription — Verbatim, including disfluencies, false starts and fillers. Moving this off the standard changes the unit cost directly.
- Timestamps — Utterance level by default; word level on request. Moving this off the standard changes the unit cost directly.
- Tagging — Noise, overlap, unintelligible, foreign-language and code-switch tags. Moving this off the standard changes the unit cost directly.
- QA threshold and rework policy: a higher accept bar means more re-collection, which is priced in rather than absorbed later
Where the cost sits in asr datasets
Transcribers are native speakers of the target variety, not of a related standard language.
Turnaround: Transcription adds roughly 1-2 weeks per 100 hours after recording, per language.
Indicative rate bands
| Scope | Unit | Indicative band |
|---|---|---|
| Widely spoken language, standard spec | per delivered hour | ₹3,200 – ₹6,500 |
| Regional language, standard spec | per delivered hour | ₹4,200 – ₹8,200 |
| Low-resource language or narrow cohort | per delivered hour | ₹5,500 – ₹11,000 |
| Verbatim transcription add-on | per audio hour | ₹900 – ₹2,400 |
| Pilot batch (10-20 hours) | fixed | Quoted separately, credited against the full run |
These are bands, not a price list. Two projects with the same hour count can differ by 3x on cohort design alone, which is why every number here is replaced by a fixed figure once we see your specification.
What the price includes
- Audio plus aligned transcripts (JSON/TSV, or your schema)
- Style guide as delivered documentation
- Inter-annotator agreement report
- Speaker-disjoint train/dev/test splits
How the quote is built
- Style guide authored per language, covering numerals, loanwords, script and disfluency rules
- Transcriber calibration round with inter-annotator agreement measurement
- First-pass transcription
- Second-pass native review
- Automated consistency checks against the style guide
- Split generation with speaker-disjoint partitions
Where budgets get wasted
The most expensive mistake in a data purchase is buying volume before the specification is settled. Re-collecting 200 hours because the noise profile did not match deployment costs more than the entire QA layer would have.
Agreement is measured, not assumed. We report word-level agreement on a held-out sample so you can judge label quality before training on it.
Best suited for
- ASR training
- ASR fine-tuning
- Domain adaptation
- WER benchmarking
Frequently asked
How much does asr datasets cost per hour in India?
Most programmes land between ₹3,200 and ₹11,000 per delivered unit depending on language tier, cohort narrowness, recording condition and annotation depth. A written specification converts that band into a single fixed number.
Is there a minimum order?
Around 50 hours or equivalent per language for a production run. Pilots of 10-20 hours are accepted when they lead into a larger build, since most setup cost sits in specification and recruitment.
Do you charge for rejected audio?
No. You pay for delivered units that pass the agreed QA threshold. Rejected material is re-collected at our cost.
Can the price be fixed rather than estimated?
Yes. Once the specification is signed the price is fixed against it. Changes to quotas, conditions or annotation depth are re-quoted before work continues.
Related pages
Get a fixed price for asr datasets
Send the volume, languages and quality bar you have in mind and you get a scoped price within a working day.