Buyer's guide
How much does egocentric video data cost? What drives the price per hour
Egocentric video data is priced per approved hour, and the rate is set by eight drivers: domain, recruiting difficulty, exclusivity, annotation, QA and rejection, metadata, consent paperwork and delivery. There is no single list price because two briefs rarely share the same mix. This guide explains what moves the number, why the approved hour is the right unit, and what to send to get a quote within one business day.
Short answer
You pay per approved hour: footage that fails human review is never paid to the contributor and never billed to you. The price per hour rises with harder domains, harder-to-recruit contributors, exclusivity, deeper annotation and heavier compliance paperwork. It falls when you buy into a domain that is already running and accept the standard metadata and delivery. Dovra supplies first-person video of real work at scale: about 5,000 hours per 4-week cycle, with consent and provenance per clip.
Why is egocentric video priced per approved hour?
Because the approved hour is the unit your model consumes. Clips of real work run anywhere from 1 to 30 minutes, so pricing per clip rewards short clips and punishes complete tasks. Pricing per recorded hour means you pay for footage a reviewer will later throw out. Per approved hour ties the invoice to usable training time: a human rates every session for quality, variety and task match before it counts, and only what passes is paid and billed. In Dovra's network the dominant rejection reason is clarity (light and phone position), which is exactly the footage you would not want to pay for.
Which domain are you buying?
Domain is the first driver because it sets who has to be recruited and where. A commercial kitchen, a garage or a warehouse is a workplace with an owner who must agree in writing, staff who must sign a contributor agreement, and tasks that only happen during a real shift. Domains that are already running cost less to serve than a new one: Dovra can deliver first clips in 72 hours in a covered domain, while a new domain takes two to three weeks to stand up, and that set-up effort is reflected in the hour rate.
How hard are the contributors to recruit?
Recruiting difficulty is the second driver. Footage of a common trade in a large city is easier to source than a specialized task in a specific region. The country mix matters too: Dovra's network spans 32 countries and is deepest in Latin America, so briefs concentrated there are served from standing capacity, while thin coverage elsewhere means slower ramp and higher cost per hour. If your spec allows several countries and scene types, you widen the pool and usually lower the rate.
Do you need exclusivity?
Exclusivity raises the price because it removes footage from the pool that could otherwise be licensed again. Most teams do not need it: what they need is a license that explicitly covers training and evaluation, with provenance they can show legal. If you do need exclusive capture, say so in the brief, because it changes how the collection is run, not just how it is priced.
How much annotation do you want?
Every Dovra clip ships labeled by domain and task from a fixed taxonomy, with scene type, country, device model and OS, duration, timestamp, session ID and a human quality rating. That baseline is included. Step segmentation, hand-object interaction tags, affordance or failure-mode marking, or mapping to your own taxonomy are annotation on request, priced per hour on top of capture. The more labeling you push to the vendor, the higher the hour rate, so decide which labels you need at training time and which you can derive yourself.
What does QA and rejection cost?
A human QA gate costs the vendor money and saves you more. Reviewing every session, enforcing exclusions and rejecting padded or purposeless footage is work, and it is built into the approved-hour rate. The alternative, a cheaper recorded-hour rate with no gate, moves that cost to your side: you pay for the footage, then pay again to filter it. Ask any vendor for rejection rates and the top reasons; a pipeline that rejects nothing is not cheaper, it is unfiltered.
How much consent paperwork and delivery do you need?
Compliance and delivery are the last two drivers. A defensible dataset needs a signed agreement per contributor that names AI training, written permission per business, a payment record per approved hour, a training license and a provenance document that ties each clip to who recorded it, where, when and on which device. Dovra includes that chain of custody. Delivery is an encrypted transfer to your bucket with a manifest and dataset card, in formats matched to your ingestion; unusual formats or custom documentation add effort and show up in the quote.
What drives the price up or down?
| Driver | Raises the rate | Lowers the rate |
|---|---|---|
| Domain | New domain, set up from zero | Domain already running (first clips in 72h) |
| Recruiting | Rare task, single region | Common trade, several countries allowed |
| Exclusivity | Exclusive capture | Non-exclusive license |
| Annotation | Step segmentation, your taxonomy | Standard domain and task labels |
| QA and rejection | Strict custom acceptance rules | Standard three-check review |
| Metadata | Extra fields beyond the standard set | Standard per-clip metadata |
| Consent paperwork | Custom agreements and audits | Standard chain-of-custody pack |
| Delivery | Custom formats and documentation | Encrypted transfer, manifest, dataset card |
What should you send to get a quote?
Send the domains and tasks you are training on, the scene types, the country mix, the minimum clip length, what must never be captured, the annotation you need and a rough number of hours. With that, Dovra replies within one business day with a per-approved-hour quote and a free sample pack that carries the same metadata and consent record as a full delivery, so you can check the footage before you commit. For a structured list of what else to check, see the 10 questions to ask a vendor, and for the basics, what egocentric video data is.
Dovra is backed by Simera, a remote talent company for Latin America. Read why Simera backs Dovra.
Get a price for your brief
Tell us the domains, scenes, country mix and roughly how many hours you need. We reply within one business day with a per-approved-hour quote and a free sample pack.
Request a quote See dataset coverageFAQ
What is the egocentric video data cost per hour?
It depends on the brief. Egocentric video is priced per approved hour, and the rate moves with the domain, how hard contributors are to recruit, exclusivity, annotation, QA and rejection rates, metadata depth, consent paperwork and delivery. Dovra quotes per approved hour within one business day.
Why is egocentric video priced per approved hour instead of per clip?
Because an hour of approved footage is the unit your model actually consumes. Clips vary from 1 to 30 minutes, so per-clip pricing rewards short clips. Per approved hour ties what you pay to usable training time, after a human review has rejected what does not meet the spec.
Do you pay for rejected footage?
No. In Dovra's pipeline every session is rated by a human before it counts. Footage that fails the quality, variety or task-match check is never paid to the contributor and never billed to the buyer.
Is annotation included in the price per hour?
Every clip ships labeled by domain and task from a fixed taxonomy. Deeper annotation such as step segmentation, hand-object interaction tags or your own taxonomy is on request and priced per hour on top of capture.
How do I get a quote?
Send the domains, tasks, scenes, country mix and rough volume through the request form at dovra.ai/#offer. Dovra replies within one business day with a per-approved-hour quote and a free sample pack carrying the same metadata as a full delivery.
Bottom line
There is no list price for egocentric video, but there is a clear unit: the approved hour, with rejected footage never billed. Domain, recruiting, exclusivity, annotation, QA, metadata, paperwork and delivery set where your rate lands. Dovra supplies first-person video of real work at scale: about 5,000 hours per 4-week cycle, with consent and provenance per clip. Send your brief and get a quote within one business day.