Indri Labs vs Defined.ai: buy a dataset that exists, or source the one you need.
Defined.ai runs a marketplace of 700+ ready-to-use AI training datasets across speech, text, image, video and multimodal data, plus custom collection, labeling and evaluation through its contributor crowd. Indri Labs is a sourcing service, not a catalog: you describe the data you need, and we find the companies that hold it, license it, record what doesn't exist yet and deliver one merged dataset. Choose Defined.ai when a dataset that already exists will do; choose Indri Labs when the data sits inside companies or has to be recorded to your spec.
Updated · Indri Labs
What is Defined.ai?
Defined.ai was founded in 2015 by Daniela Braga as DefinedCrowd and took its current name in 2021. It is headquartered in Seattle, with offices in Lisbon, Porto and Tokyo. It calls itself the leading marketplace for ethical AI training data.
Its marketplace, which it describes as the largest selection of ready-to-use AI training datasets, lists 700+ datasets across speech, text, image, video and multimodal data. Buyers can filter by language, domain and format, get free samples, request quotes and receive data by file transfer or API. Third parties can also sell datasets there; Defined.ai says partner datasets on its marketplace grew 1,200% in 2025.
Alongside the catalog, Defined.ai offers custom data collection, annotation, transcription and evaluation, including red teaming. It says these run through a community of 1.6M+ vetted contributors across 150+ markets, covering 500+ languages, dialects and locales. It lists ISO 27001, ISO 27701 and ISO 42001 certification and GDPR and HIPAA compliance, and says all of its data is sourced with informed consent and fair compensation.
What is Indri Labs?
Indri Labs sources proprietary training data for AI labs as a service. There is no catalog. You describe the need: modality, volume, coverage across regions, languages and edge cases, format, rights and deadline. We scope it for free and send a one-page brief that names and prices every source, includes a sample from each, gives the total data cost and shows our fee on its own line. You can walk away owing nothing.
If you approve, you sign one contract and pay one invoice, with us. We negotiate a license with every seller, usually companies that have never sold their data, and record what doesn't exist yet, such as first-person video on smart glasses and head cameras filmed to spec. You receive one merged dataset with duplicates and personal data removed, checked against your spec, with the source and license for every record, in your format.
How do Defined.ai and Indri Labs compare?
The table sets out the two models side by side. Statements about Defined.ai come from its own website and press material as of October 2026, listed under sources.
| Aspect | Defined.ai | Indri Labs |
|---|---|---|
| Model | Dataset marketplace plus custom data services | Sourcing service: scopes, licenses and records data to spec |
| What you buy | A license to an existing dataset, or a custom project | One merged dataset built from the sources named in your brief |
| Data types | Speech, text, image, video, multimodal | First-person video, operational logs, calls and meetings, documents; sensor, robotics and clinical records coming soon |
| Custom collection | Yes, through a crowd it puts at 1.6M+ contributors in 150+ markets | Yes, first-person video on smart glasses and head cameras filmed to spec |
| Pricing visibility | No public price list; free samples and quotes on request | No public price list; free brief prices every source, with the fee on its own line |
| Contract | With Defined.ai | One contract and one invoice with Indri Labs, covering every seller |
| Provenance | Defined.ai says all data is sourced with informed consent and fair compensation | Source and license recorded for every record |
| Best for | Teams that need a dataset this week and can accept one that already exists | Teams that need data no catalog lists, from companies that have never sold it |
What is the main difference between a dataset marketplace and a sourcing service?
A marketplace sells what sellers have already prepared. That makes it fast: you can browse, sample and license a speech or image dataset without a long negotiation. The limit is that you choose from what is listed, and the closest match may not cover the exact languages, settings or edge cases you need.
A sourcing service starts from your spec instead of a shelf. Much useful training data, such as operational logs, recorded calls and internal documents, sits inside companies that have never offered it for sale. Getting it means finding those companies, agreeing a license with each and cleaning the results into one dataset. That takes longer than a purchase from a catalog, but it reaches data that no catalog holds.
Defined.ai covers part of this ground too, through custom collection with its crowd. The difference is the source: crowd collection creates new data with paid contributors, while Indri Labs licenses existing records from the companies that hold them and films first-person video to spec where nothing exists.
When is Defined.ai the better choice?
- A dataset in its catalog matches your need and you want it this week
- You want to filter by language, domain and format and test free samples before buying
- You need speech or text data across many languages and locales
- You need annotation, transcription, evaluation or red teaming run by a large contributor community
- You want API delivery of off-the-shelf data
When is Indri Labs the better choice?
- The data you need is held inside companies and not listed in any marketplace
- You need first-person video recorded on smart glasses or head cameras to your spec
- You want several sources merged into one dataset under one contract and one invoice
- You need the source and license for every record, and the data cost and fee shown separately
- You need operational logs, real calls and meetings with transcripts, or company documents
Many labs use both: a catalog for data that exists and a sourcing service for the gaps. For the wider field, see AI training data providers compared, and for first-person video, egocentric video data. If your company holds data that AI labs could use, Defined.ai lets third parties list datasets in its marketplace, and Indri Labs can also help you sell your data to labs.
Questions, answered.
What is Defined.ai?
Defined.ai, founded in 2015 by Daniela Braga and headquartered in Seattle, runs a marketplace for AI training data. It says the marketplace holds 700+ ready-to-use datasets across speech, text, image, video and multimodal data, with free samples and delivery by file transfer or API. It also offers custom data collection, annotation and evaluation.
Is Indri Labs a Defined.ai alternative?
Yes, for a different kind of need. Defined.ai sells datasets that already exist in its catalog. Indri Labs is a sourcing service: it takes a description of the data you need, finds and licenses it from the companies that hold it, records what doesn't exist yet and delivers one merged dataset under one contract.
Where can I buy speech datasets?
Defined.ai lists speech datasets in its marketplace, filterable by language, domain and format, with free samples. If you need recordings of real calls and meetings held by companies, with transcripts and a license for every record, Indri Labs can source them on request.
Which is faster, Defined.ai or Indri Labs?
A catalog is faster when the dataset you need already exists, so Defined.ai is usually the quicker route for off-the-shelf data. Indri Labs takes longer because it negotiates with data owners and may record new data, but it reaches data that no marketplace lists.
Does Defined.ai or Indri Labs publish prices?
Neither publishes a standard price list. Defined.ai offers free samples and quotes on request. Indri Labs scopes your need for free and sends a brief that names and prices every source, gives the total data cost and shows its own fee on a separate line.