01
Signed consent.
Every contributor signs before anything is recorded.
Training
data,
made
to
order.
Speech, video and expert work, collected to your spec and delivered with full training rights.
What we supply
Conversations and read speech, in the languages and accents you name.
- You specify
- Language
- Accent
- Age range
- Device
- Setting
First-person and multi-angle footage of people doing real tasks.
- You specify
- Task
- Camera angle
- Environment
- Length
Reasoning and dialogue written by practitioners in your field.
- You specify
- Field
- Credentials
- Format
- Rubric
Rankings, rubric scores and held-out test sets.
- You specify
- Rater background
- Rubric
- Agreement level
Photos and documents, labeled to your schema.
- You specify
- Subject
- Conditions
- Label schema
How it works.
01You set the spec.
Who, what, how much, and what counts as good.
02We collect.
We recruit contributors who match your spec and run the collection.
03We check.
Every file is reviewed against your acceptance criteria.
04You receive.
Files in your format, with the license attached.
What you receive
Every delivery, complete.
Data, paperwork and proof arrive together.
The files
In the format, naming and folder structure you specify.
Annotations
Transcripts, labels and scores, to your schema.
Metadata
Language, device, setting and any field you ask for, per file.
QA report
What was checked, what was rejected and why.
License
One agreement covering training, evaluation and commercial use, with consent records attached.
Consent and rights
Rights you can show to legal.
02
Full training rights.
Train, fine-tune, evaluate and ship. Exclusive if you need it.
03
Traceable to the file.
Each file links to the person and the agreement behind it.
Getting started
Start with a pilot.
First
Pilot
A small batch built to your spec. Judge it on your own criteria before you commit.
Then
Production
The full volume, delivered in batches to the agreed spec.
After
Ongoing
New batches on a schedule, as your models and needs change.
Fields we take briefs in
- Low-resource languages
- Accented speech
- Skilled trades
- Clinical reasoning
- Legal analysis
- Software engineering
- Robotics
- Finance
- Logistics
- Scientific writing
Send a spec or a rough description. We reply with questions or a plan, then build a pilot batch before you commit to volume.
It can be. Data collected for your brief can be yours alone, and you choose an exclusive or a non-exclusive license.
From contributors we recruit for your project and from owners who license their own data through us, each under a signed agreement. Contributors are paid a share when their data is licensed. Nothing is scraped.
Against acceptance criteria agreed with you before collection starts. Files that miss them are replaced.
A license to train, fine-tune, evaluate and deploy commercially.
Per hour, per item or per task. It depends on the type of data, the volume and how specialized the work is.
Yes. Send us your files and we transcribe, label or evaluate them to your schema.
Yes. We sign an NDA before you share anything, and your brief is never shown to another client.
Work with us
Two ways to work with us.
For AI teams
Source training data.
Custom speech, video, expert text, evaluation and image datasets, delivered to your specification and licensed for training.
Request a proposal
For data owners
License your data.
License the audio, video, documents or records you hold to AI developers. We review the collection, confirm the rights and return with proposed terms. You are paid a share when your data is licensed.
Submit your data for reviewWe license only data with clear ownership and documented rights.
Tell us what your models need.
Get startedHold data that AI developers could use? Explore a data partnership