Speech data pricing per hour of audio
Every dataset in our catalog has a published price per hour of audio, from $60 to $95 depending on the language. One hour is one hour of recorded two-person conversation with dual-channel audio, time-aligned and human-verified transcripts, speaker metadata, a commercial license and consent documentation for every speaker.
Prices by region
| Region | Per hour | Datasets | Hours |
|---|---|---|---|
| Europe | $75 to $95 | 22 | 21,500 |
| South Asia | $65 to $75 | 10 | 8,500 |
| Southeast Asia | $60 to $75 | 6 | 12,000 |
| East Asia | $70 to $95 | 3 | 2,500 |
| Middle East | $65 to $95 | 7 | 8,000 |
| Africa | $60 to $65 | 8 | 14,000 |
| North America | $90 | 1 | 1,500 |
| Latin America | $65 | 3 | 5,000 |
| All regions | $60 to $95 | 60 | 73,000 |
List prices in US dollars. Each dataset page shows its own price. Volume and multi-language discounts on request.
What a dataset costs
Hours times the published price, before any volume or multi-language discount.
500 hours of German
German: 500 h × $95
The full Hindi dataset
Hindi: 2,000 h × $65
Three Arabic variants, 500 hours each
Egyptian: 500 h × $65
Gulf: 500 h × $90
Levantine: 500 h × $65
What changes the price of custom collection
Custom collection is quoted per project, because the cost depends on the spec. Send yours and we reply within one business day with feasibility and a quote.
Language and recruitment
How hard it is to find and vet native speakers of the language, dialect or accent you need.
Speakers and quotas
How many speakers you need, and how tightly they must match quotas for age, gender, region or accent.
Scripted or spontaneous
Read prompts are simpler to record and align. Natural conversation needs paired speakers and more transcription work.
Devices and environments
Specific phones, headsets, far-field microphones, in-car audio or background noise add setup and QA.
Transcription and labels
Verbatim conventions, disfluency, overlap and code-switch tags, or extra labels add annotation time.
Timeline
A tighter deadline means recruiting and recording more sessions in parallel.
Call-center recordings and audio annotation are also quoted per project.
Pricing questions
How much does speech training data cost per hour?
SpeechData.ai's off-the-shelf conversational datasets cost $60 to $95 per hour of audio, depending on the language. Each dataset page shows its price. Custom collection is quoted per project.
Are there volume discounts?
Yes, on request, for full-dataset and multi-language purchases. The prices on this page and on each dataset page are list prices.
What is included in the price?
Dual-channel audio of spontaneous two-person conversations recorded in-country, time-aligned and human-verified transcripts, speaker metadata (age, gender, region and recording device), a commercial license for training and evaluating ASR, TTS and other voice AI models, and consent documentation for every speaker.
Can I get a sample before buying?
Yes. Free sample audio, matching transcripts and the metadata schema are sent on request to a company email.
How is custom speech data collection priced?
Per project, based on the language, speaker count and quotas, speech type, devices and environments, transcription depth and timeline. Send your spec and we reply within one business day with feasibility and a quote.
Check the data before you buy
Free sample audio, matching transcripts and the metadata schema, sent to your company email.