Voice Data Supplier
For studios, agencies, and teams that can produce voice recordings to a specification: what buyers check, how repeat work actually arrives, and what ends a supplier relationship.
-
A supplier is judged on repeatability
Buyers are not looking for the best single delivery. They are looking for someone who will deliver the same quality next time without supervision.
-
Documentation is the differentiator
Two suppliers with equal recording quality are separated by who can hand over consent records and metadata without being chased.
-
Repeat work arrives quietly
It comes as a short message asking for the same thing again, not as a new tender. Suppliers who have moved on by then lose it.
-
Relationships end on process, not on quality
The usual cause is a missed deadline or an undocumented contributor, not bad audio.
What a voice data supplier is
A voice data supplier is an organization that can produce speech recordings to someone else's specification, repeatedly, with documentation. It differs from an individual contributor in three ways: capacity, process, and paperwork. A studio with several recording rooms, a recruitment pipeline, and a signed-consent workflow is a supplier. A person with a good microphone and a voice is a participant in someone else's project.
The distinction matters because data is bought against a specification, not browsed. Buyers do not purchase "a voice"; they purchase a volume of finished, documented audio that meets a written specification on a date. The supplier's product is the process that produces that reliably.
How a buyer evaluates a new supplier
A buyer adding an unknown supplier is not comparing catalogues, because there is no catalogue. The assessment is a short sequence, and any step can end the conversation before audio is ever discussed.
- A capability statement first: languages, recording conditions, annotation depth, monthly capacity in hours. Vague statements do not proceed, because there is nothing to match against a live requirement.
- Then a documentation request: the speaker release form, the retention policy, and the speaker manifest from a previous project. This is the step where most new suppliers are eliminated, and it is rarely about the quality of their audio.
- Then a paid pilot batch, usually a few hours produced to the real specification and reviewed against the real criteria. Suppliers who treat the pilot as a formality tend to fail it.
- Then a process check on the pilot, not the recordings, but how they were made: who recruited, who recorded, who annotated, who holds the consent.
- Then a first real order. Volume follows clean delivery rather than arriving with it, which is the opposite of how most new suppliers expect the sequence to run.
How repeat work actually arrives
It does not arrive from a registry. There is no list a supplier joins that causes buyers to call, and no public portal at any major lab where voice data can be uploaded for payment. OpenAI, Google, and Anthropic buy through data teams and vendors, not from submissions. What exists is matching: a live requirement appears, someone who holds it looks for producers who can meet it, and suppliers with a documented capability get the specification.
That is the function a broker performs, and it is why the useful first move for a new supplier is a specific capability statement rather than a sales pitch. Once a supplier has delivered one project cleanly, the second arrives with less friction: repeat work in this market is earned per project, not granted once.
Payment follows acceptance on every project: the buyer's QA reviews the delivery against criteria set at the start, and the batch is paid when it passes. Suppliers who price to absorb one round of re-work survive; suppliers who treat delivery as the end of the obligation do not get a second project.
What ends a supplier relationship
Four things, in the order we see them. A rights gap. A consent form that turns out not to cover the use the data was sold for. It is the most common failure and the most fatal: a supplier who cannot document consent cannot be paid, whatever the recording sounds like. Then reused speakers presented as a larger pool than they are, which buyers detect through the speaker manifest. Quality drift after the pilot, usually from quietly moving work to a cheaper subcontractor. And missed dates without warning, which is survivable once and never twice.
Two categories end a relationship before it starts, for us and for the buyers we work with: recorded telephone calls, where the other party never consented, and medical or clinical recordings. Suppliers whose main asset is either should not spend time on this market.
What to have ready before your first buyer conversation
Most supplier conversations stall because one side cannot answer a basic question. Having these ready turns a slow exchange into a short one.
It is worth writing them down. A supplier who can send a one-page summary immediately is treated differently from one who promises to check and reply later.
- A capability statement. Which languages, which dialects, which recording conditions, and how many hours per month you can genuinely hold.
- A consent template. The release form you use, and a plain description of what contributors are agreeing to.
- A sample with metadata. A short recording that comes with its recording conditions, speaker profile and annotation format.
- A quality control description. What you check, how often, and what you do when something fails.
- Two references or past deliveries. Evidence that a buyer has accepted your work before, even at small scale.
- Your commercial terms. Rate structure, payment terms and lead time, stated up front rather than negotiated from scratch each time.
Questions we get asked
Do you require exclusivity from suppliers?
No. You can work with other buyers and brokers. The only restriction we ask for is that data produced against a specific buyer's order is not resold to anyone else.
How large does a supplier need to be?
There is no size threshold. A small studio with genuine native-speaker access in an uncommon language is more valuable than a large firm that only covers common ones. What matters is that your stated capacity is real.
What does a first project look like?
A specification, a quote, and a pilot batch reviewed against agreed criteria. The pilot protects both sides: it proves your output meets the bar before a full run is committed, and it is paid work within the project rather than free inventory.
Keep reading
-
Sell AI Training Data
If you have recording capability, annotation capacity, or language access that AI teams need, we want to hear from you.
-
Sell Data to AI Companies
How AI companies actually buy data, the routes a supplier can realistically use, and the reasons most first approaches fail before anyone looks at the data.
-
Sell Data to AI Labs
Labs buy differently from product companies: narrower specifications, smaller volumes, and a higher bar on provenance. What a supplier should expect from that process.
-
How we place your data
Who buys the data you supply, and what we handle between the two of you.
Offer your data capacity
Tell us what you can produce, in which languages, and how much per month. If there is a buyer for it, we come back with a real scope, and we carry the contract, the compliance paperwork, and the delivery risk, not you.
- You record; we handle the buyer, the contract, and the paperwork.
- Paid per accepted batch. No listing fee, no exclusivity clause.
- We do not buy medical or clinical data, or intercepted call recordings.
- You tell us your capacity What you can produce, which languages, how much per month.
- We reply within two business days Whether we have a buyer for it, and what that buyer would need.
- You deliver a pilot batch A small run first, so both sides see the quality before a full order.
- We handle the buyer and the paperwork You deliver batches and get paid per accepted batch.