Defined.ai

1. Custom collection across audio, image, video, text and multimodal: everyday objects, facial expressions, gestures; egocentric POV, cinematic, action-sequence and behavioural clips; sensor-based robotics data; 1.6M+ experts, 150+ countries, 500+ languages; 'delivered quickly and competitively'; fully consented and copyright-cleared; multi-layer validation and real-time dashboards (redirect target of the old /ai-data-collection-generation URL) defined.ai
2. Custom collection scoped to accuracy, demographic, language and compliance requirements; 700+ marketplace datasets; Crowd-as-a-Service; 150+ markets; legal documents list defined.ai
3. Contributor recruitment and NDA/MSA; scripted speech recorded on-device with speaker and device information; dialogue and IVR via telephony; standard licence covers all datasets and permits commercialised models; custom collection can be ordered; USD ACH payment, POs and SOWs; no refunds; custom orders take longer; academic due diligence defined.ai
4. Data License Agreement: non-exclusive, non-sublicensable licence; no sale or sublicensing of the data (§1(b)); models trained, tested or benchmarked on the data may be commercialised (§1(c)); Licensor owns the data (§4); export control (§9(g)) defined.ai
5. Website terms: robots and automated access to the website are prohibited (applies to the website, not the API) defined.ai
6. Public API: Enterprise account via sales; OAuth2 client credentials with PublicAPIv2 scope; base URL; clone, upload, status and deliverables endpoints; workflow templates developers.definedcrowd.com
7. OpenAPI 3.0.1 document with endpoints, Bearer JWT header and error codes api.production-na01.definedcrowd.com
8. OAuth metadata: token endpoint, PublicAPIv2 scope, client_credentials grant, PKCE plain/S256, no registration endpoint auth.definedcrowd.com
9. Robotics sensor collection: 225 hours of synchronized POV video, gaze, depth, motion-capture and force data in two weeks within two months; drop-frame KPI; UTC time and location metadata as project requirements defined.ai
10. Crowd-as-a-Service via the Neevo platform; scripted speech recorded on mobile devices and validated by multiple annotators defined.ai
11. ISO 27001, 27701 and 42001 certification; consent-based sourcing with no scraping; multi-pass QA and verified annotators defined.ai
12. No Defined.ai or DefinedCrowd server in the MCP Registry registry.modelcontextprotocol.io
13. DLA §1(c) internal use, §8 termination and deletion defined.ai

About. AI training-data company (formerly DefinedCrowd) that, besides its dataset marketplace, runs custom collections with a vetted global crowd through its Neevo platform: images (everyday objects, facial expressions, gestures), video (egocentric POV, action sequences, behavioural clips), speech recordings, and wearable sensor data for robotics. Collections are scoped to demographic, language and compliance requirements. Sales-led and quote-only; an enterprise REST API (OAuth2 client credentials) lets account holders clone project templates, upload prompts, track status and download deliverables.

Price. from_usd 0 means unknown, not free. No prices for custom collection; volume discounts by quotation. Marketplace payments are USD by ACH bank transfer, with purchase orders and SOWs on request; no refunds.

Coverage. Defined.ai cites 1.6M+ contributors in 150+ countries (also '150+ markets') and 500+ languages and locales across all services; how many capture images or video, and where, is not published. The public API's language list names locales, not collection countries. An in-house recording studio operates in Lisbon.

Timing. No turnaround published; the site says only 'delivered quickly and competitively', and the FAQ says custom orders take longer.

Licence. The FAQ says all Defined.ai datasets are covered by its standard Data License Agreement (effective 30 Nov 2022). That licence is non-exclusive, non-sublicensable and non-transferable; §1(c) permits internal use and commercial exploitation of models trained, tested or benchmarked on the data, provided the data is not made available and the model does not output recognisable contributor content (e.g. TTS voices, avatars). §1(b)(ii) and §1(c) forbid selling, sublicensing or otherwise commercialising the data itself, and §4 says Defined.ai owns all rights in the data. The site says every dataset is 'fully consented, copyright-cleared, and privacy-compliant' (GDPR, HIPAA). Use is for internal business operations only (§1(c)), and on termination (60 days' notice, or breach) the licensee must delete or return all copies of the data (§8), despite the FAQ calling the licence perpetual. Whether a custom collection contract changes these terms, including ownership or exclusivity, is not published.