Data Data Partnerships Data Value Estimator Referrals Design Search Lead generation Data guides Blog Get an offer →
Theme
← Company data sourcesData source · Phone and call recordings

How to find and export your Twilio data for AI training

Twilio shows how much recording data you have in the recordingstorage usage category, which reports stored recordings and stored minutes (divide by 60 for hours) and can be read in the Console or through the Usage Records API, and a developer exports the full history with the Recordings API, up to 1,000 results per page. Recordings stay in the account until deleted, but Voice Insights shows only 7 days of calls, or 30 with Advanced Features. Below, you'll find what Twilio data is and why AI labs and data vendors buy it, where to read your total, what to know before you apply, how the export works, and what stays private.

Where the total is
recordingstorage usage record; minutes divided by 60
Who can read it
Console access or API credentials; no per-seat license
Recordings download as
WAV at 128 kbps or MP3 at 32 kbps
Retention
Kept until deleted; details stay 40 days after

Key takeaways

  • Twilio reports stored recordings and stored minutes in the recordingstorage usage category; divide the minutes by 60 to get the hours to report.
  • A full export is an API job: a developer lists recordings by date created, up to 1,000 per page, and downloads each one as a WAV (128 kbps) or MP3 (32 kbps) file.
  • Recordings stay in the account until you delete them, and a deleted recording's details remain for 40 days; Voice Insights shows only 7 days of calls, or 30 with Advanced Features.
  • You don't need to export anything to apply: have your hours, years of history and the number of people taking calls ready, and the data vendor anonymizes the records before any AI lab sees them.

01

What is Twilio data, and why do AI labs and data vendors buy it?

The short answer

Twilio data, for this purpose, is the call recordings a company's phone applications have made through Twilio, along with the records that describe each call. AI labs (companies that build and train AI models) and data vendors (companies that anonymize business records and package them for those labs) buy an anonymized copy because a recorded support or dispatch call captures real people explaining a problem and a company's staff resolving it.

What a Twilio account records

Twilio is a service that software developers use to add phone calls and text messages to their own applications. A company might build a support line, a dispatch desk or an appointment line on Twilio, and the account behind that application can record the calls. Each recording is an audio file, and Twilio stores it in the company's account until the company deletes it.

Next to the audio, Twilio keeps metadata for each recording, which is the information that describes it, such as when it was made and how long it ran. If transcription was turned on, a written version of the call exists as well. Twilio bills for it separately.

Why recorded calls teach an AI model something

An AI model learns by studying a very large number of examples and finding the patterns in them, and those examples are called training data. Recordings from a company's own call flows are especially useful because the company designed the process the call follows. The caller reaches a menu, explains a request, and an agent or an automated system responds in the way the business intended.

Those recordings can also be matched to the application's own records, such as the ticket or appointment that the call created. That match shows what was said and what happened afterward.

AI labs now want to train AI agents, software that carries out a job step by step, and an agent learns a job from records of that job being done. A data vendor can turn anonymized call recordings into worked examples that an agent studies like an answer key, or build a practice copy of a call flow in which the agent is scored on each attempt. The full mechanism is in AI training data: why AI companies want your company's data.

How Twilio fits into a dataset

Twilio recordings are usually one part of a company's dataset, next to email, documents and ticket histories. Companies are paid $100K to $2M or more per dataset, according to micro1, the data vendor we typically refer companies to. The price covers the whole dataset, not each app on its own, and what your data is worth explains what moves it.

02

Where does Twilio show how much data you have?

The short answer

Twilio tracks recordings in usage records. The recordingstorage category reports how many recordings are stored and for how many minutes in total, and dividing the minutes by 60 gives the hours to report. Anyone with access to the Twilio account can see usage and call recordings in the Console, and a developer can read all-time totals through the Usage Records API.

How to read your recorded minutes

The steps below read the size of your recordings from Twilio's usage data, which Twilio documents on its Usage Records page.

  1. Sign in to the Twilio Console and open Call Recordings to browse the recordings in the account. In the legacy Console the same list is called Recording Logs.
  2. Check the usage figures for the call recordings and recording storage categories. For recording storage, the Count is the number of stored recordings and the Usage is the number of stored minutes.
  3. For all-time totals, ask a developer to read them from the Usage Records API. By default the API returns one record per category covering everything the account has ever used.
  4. Divide the stored minutes by 60 to get hours.

A usage record has three amounts: Usage (for recordings, the minutes), Count (the number of recordings) and Price. The recordings category counts recordings made, while recordingstorage counts those still held in the account, and the second is the one that matters for what you can license.

Reading the total through the API

An API (application programming interface) is a way for one program to ask another for data directly. A developer with the account SID (the account's identifying number, which starts with AC) and its auth token (the account's secret key) asks for the all-time recording storage record like this:

curl -G https://api.twilio.com/2010-04-01/Accounts/ACXXXXXXXX/Usage/Records/AllTime.json \
     -d Category=recordingstorage \
     -u ACXXXXXXXX:YOUR_AUTH_TOKEN

The response lists a count (stored recordings) and a usage figure (stored minutes). The developer replaces the placeholders with your account SID and token.

What Voice Insights shows, and why it is not enough

The Console also has Monitor > Insights > Voice for call counts. It is a monitoring tool for recent calls, not a history, because since June 1, 2024 it shows 7 days of call data, or 30 days when the Advanced Features add-on is switched on. Twilio announced the change in its changelog. The Advanced Features setting is under Monitor > Insights > Voice > Settings.

HOW FAR BACK EACH VIEW REACHES The same Twilio account Voice Insights7 days Voice Insights Advanced30 days Stored recordingsUntil deleted Deleted recording's details40 days
Only the stored recordings reach back through the whole life of the account, so they are what you measure and what you export.

Who can see it

Reading usage takes either a login to the Twilio Console or the account's API credentials. Twilio has no per-seat license, so anyone you have given one of those can read usage.

03

What should you know about your Twilio data before you apply?

The short answer

Have three facts ready: the hours of recordings stored in your Twilio account, how many years you have used it and how many people make or take calls through your application. They go into the data manifest, the written summary of which apps hold your records, how many years each covers and roughly how much. You don't need to run an export to find them.

The hours of audio

Take the stored recorded minutes from the recordingstorage usage record and divide by 60, then enter the result with hours as the unit. This counts the audio still held in the account. Recordings deleted earlier are not part of it, because they are gone.

The years of history

For years in use, take the date the Twilio account was created, or the date of the oldest recording the API returns, whichever fits how long the company has been recording calls.

The number of seats

Twilio does not license by seat, so count the agents or users who make or take calls through your application instead.

What you don't need to do first

You don't need to export anything to apply, and you don't need to clean or scrub an export. The data vendor anonymizes the records before anyone else sees them, and nothing moves until a lawyer-reviewed license is signed. The application, the intro call and each later step are laid out in how to sell your company's data.

04

How do you export your full Twilio call history?

The short answer

Exporting everything from Twilio is an API job, not a Console download. A developer asks the Recordings API for the recordings made in each date range, reads up to 1,000 results per page, and then downloads every recording as a WAV file (128 kbps) or an MP3 file (32 kbps). Nothing expires on its own, since recordings stay in the account until someone deletes them.

The export step by step

Twilio documents each call in its Recordings resource page. A developer who holds the account SID and auth token follows these steps.

  1. Request the list of recordings from the Recordings endpoint, filtering by date created so that each request covers one period (a month, for example).
  2. Set the page size to 1,000, which is the maximum, and follow the next-page link in each response until the list ends.
  3. Keep each recording's metadata (its date, duration and status) in a spreadsheet or database.
  4. Download the audio for every recording whose status is completed, using the media address that Twilio gives for it.
curl -G https://api.twilio.com/2010-04-01/Accounts/ACXXXXXXXX/Recordings.json \
     -d DateCreatedAfter=2025-01-01 -d DateCreatedBefore=2025-02-01 -d PageSize=1000 \
     -u ACXXXXXXXX:YOUR_AUTH_TOKEN

This request lists the recordings made in January 2025 on the account. The developer repeats it for each period, then fetches the audio itself.

Audio formats and credentials

By default, a recording downloads as a WAV file at 128 kbps (kilobits per second, a measure of audio quality). Adding .mp3 to the recording's address returns an MP3 file at 32 kbps, which is smaller but lower in quality. Twilio stores call and conference recordings with each side of the call on its own audio channel, and a request that adds ?RequestedChannels=2 keeps them separate in the download. The media addresses require your account credentials, so they cannot be shared as plain links.

What gets stored, and for how long

Twilio charges every month for each stored minute of recording, and a recording stays put until you delete it. Once one is deleted, its metadata remains visible in the Console and the API for 40 days, while the audio itself cannot be brought back. A recording can be fetched only while its status is completed and its media is held at Twilio.

Voice Insights data lasts only 30 days at most, so it is not a source of long call history. Transcriptions exist only for calls where they were turned on, and they are billed separately.

How long an export takes

Twilio gives no time estimate. The time depends on how many recordings the account holds, the page size and how fast the developer's script can download the audio.

Sources: Twilio documentation, Recordings resource, Usage Records resource and Upcoming Changes to Voice Insights Console Experience. Steps checked against these pages in November 2026.

05

What stays private when Twilio data is licensed?

The short answer

Because the recordings sit in your own Twilio account, the company decides which phone numbers, applications and date ranges are pulled. The data vendor replaces names and other personal details with consistent stand-ins before an AI lab sees anything. The company keeps ownership of its recordings, and the license covers defined records for a defined use.

You choose what is in scope

A developer pulls recordings by date range, so a company can license one phone line or one period and leave the rest untouched. The data vendor also deletes records too sensitive to clean safely instead of anonymizing them.

Names are replaced before a lab sees them

Anonymization means replacing every detail that identifies a person, such as a name or a phone number, with a stand-in that is the same every time it appears. The call still reads in order, so a lab can learn from it, but it no longer says who the people are. The diagram below shows the details of one call before and after.

IN THE TWILIO EXPORT One recorded call Caller: Maria Gomez From: +1 415 555 0132 To: Ridgeline HVAC dispatch line Agent: Jon Park Length: 6 min 40 sec Outcome: Technician booked Real names and a real phone number. WHAT AN AI LAB RECEIVES One recorded call Caller: Contact 57 From: Phone 57 To: Company line 2 Agent: Employee 12 Length: 6 min 40 sec Outcome: Technician booked Same call; the length and outcome stay.
The same call in both panels. The people, the number and the company line become stand-ins, and the length and outcome that show how the call went stay as they were.

The method, including what is deleted outright and how you review samples, is in how company data is anonymized for AI training.

Ownership and payment

Licensing gives the buyer limited rights to a copy, while the company keeps ownership of the originals. The terms that matter, such as exclusivity, permitted use and deletion, are covered in data licensing. micro1 pays in full within 30 days of accepting the data, usually as a single payment.

Questions

Frequently asked questions

Does Twilio delete call recordings on its own?

No. Twilio keeps recordings and bills for the storage each month until you delete them. After you delete one, its details stay visible in the Console and API for 40 days, but the audio cannot be recovered.

Why does Twilio Voice Insights show only a week of calls?

Twilio changed it on June 1, 2024: the Console now holds 7 days of call data, or 30 days if you pay for the Advanced Features add-on. It is a monitoring view, not a call history, so use your stored recordings and usage records for long-term totals.

Which file type do Twilio call recordings download as?

A WAV file at 128 kbps unless you ask otherwise. If you add .mp3 to the recording's address, you get an MP3 at 32 kbps instead.

How do I find out how many hours of recordings my Twilio account holds?

Read the recordingstorage usage record, which reports the number of stored recordings and the stored minutes, and divide the minutes by 60. The Usage Records API returns all-time totals by default.

Is there a way to download all Twilio recordings from the Console?

Not in one step. A full export is an API job: a developer lists recordings with the Recordings API and downloads each one, so plan on involving whoever built your Twilio application.

Do we need to export Twilio before we apply?

No. The application takes about two minutes and shares no data. You only need your approximate hours of recordings, your years of history and the number of people who take calls for the intro call. The full export is needed later, when the records are transferred after you sign the license.

What is Twilio data worth to AI labs and data vendors?

Call recordings are priced as part of a whole dataset rather than on their own. Companies are paid $100K to $2M or more per dataset, according to micro1, and the price depends on the years of history and how completely the records show each job. What your data is worth explains the factors.

Bottom line

Know what is in Twilio, then see what it is worth

You don't need to export anything to apply. Apply in two minutes, describe your records on the intro call, and micro1, the data vendor we typically refer companies to, prices the whole dataset and sends an offer 1–2 days later.

Not legal, tax or financial advice. Figures come from micro1's published pages and public reporting as of November 2026.