Call center speech data for voice agents and contact-center ASR
We record role-played customer service calls with native speakers playing the agent and the customer. Each call follows a scenario for your industry, is captured with one channel per speaker, and is transcribed with speaker turns and timestamps. Names and account details are fictional, so the recordings contain no real customer data.
Realistic calls, without real customer data
Scenario scripts per industry
Each call starts from a scenario: who is calling, why, and what the account shows. Examples include banking, insurance, telecom, healthcare scheduling, travel, retail and e-commerce, and utilities, or the call types your own contact center handles.
Unscripted dialogue inside the scenario
Speakers improvise rather than read lines, so calls carry the interruptions, holds, transfers and complaints that real calls contain.
Recorded like a call
Agent and customer are recorded on separate channels. Capture is narrowband telephony or wideband audio as your brief specifies, on the devices and channels you scope, such as mobile, landline or VoIP.
What a project can include
| Speakers | Native speakers playing agent and customer, recruited to your quotas for age, gender and region |
|---|---|
| Scenarios | Industry scripts agreed with you, from routine requests to escalations |
| Audio | One channel per speaker, narrowband telephony or wideband |
| Transcripts | Time-aligned and human-verified in your convention, with agent and customer turns |
| Optional labels | Intent, call outcome, sentiment, and PII entity tags on the fictional details for redaction models |
| Consent | Explicit commercial-use consent from every speaker, documented |
| Languages | Any of our 60 catalog languages and dialects, and others on request |
Built for
- Voice agents and IVR
- Contact-center speech recognition
- Call analytics and quality monitoring
- Agent assist
- Held-out evaluation sets for production calls
Catalog or custom?
Our off-the-shelf datasets are general two-person conversations, recorded with one channel per speaker. They are not customer service calls, but they can be a starting point for conversational ASR. Call center data is recorded to your spec.
Frequently asked questions
Where can I buy call center audio data to train a voice agent?
SpeechData.ai records call-center style conversational audio to order: role-played customer service calls between native speakers playing agent and customer, in your industry scenarios and target language, with one channel per speaker and time-aligned transcripts. It is recorded for each project rather than sold from stock, and you get a feasibility assessment and a quote within one business day.
Do the recordings contain real customer data?
No. The calls are role-played, and the names, account numbers and other details in each scenario are fictional. Every speaker signs explicit commercial-use consent before recording, and the consent documentation is delivered with the data.
Can you deliver 8 kHz telephony audio?
Yes. Calls can be captured as narrowband telephony audio or as wideband audio, as your brief specifies, on the devices and channels you scope, such as mobile, landline or VoIP.
Which languages can you record?
Any of the 60 languages and dialects in our catalog, and other languages on request, such as isiZulu, isiXhosa and Sesotho. We confirm recruitment feasibility for each language before quoting.
Describe the calls your model needs to handle
Share your languages, industries, call types, hours and audio channel. We’ll come back within one business day with a feasibility assessment and a quote.