Everything you need to record and deliver footage that passes QC the first time. Ten minutes of reading that pays for itself on your first clip.
Two clips from a real, paid Firstly Labs contributor. Notice the chest-mounted POV, both hands in frame, no faces, and the claim code note sitting on the work surface — not held up to the camera. Read the rules below to see how these were built.
Your footage teaches robots and physical-AI systems how humans actually do everyday work — how hands grip, move, adjust, and recover. First-person video from real homes is the training data the entire field is short on. That's why it pays.
A real, lived-in home — yours, or a friend's or family member's with their permission. Real homes are the point: genuine clutter, genuine appliances, genuine life.
Your phone records from your point of view while both hands do the work. A chest mount or head strap (GoPro-style phone clamp, $10–20 anywhere online) is the standard rig. Smart glasses work for shorter tasks if you own them.
| Parameter | Requirement | |
|---|---|---|
| Resolution | 1920×1080 (Full HD) minimum — 1080p is the acceptance floor | REQUIRED |
| Frame rate | 30 FPS minimum · 60 FPS earns the top rate tier | REQUIRED |
| Orientation | Landscape (horizontal). Never vertical. | REJECTED IF VERTICAL |
| Field of view | Wide — 0.5× or 0.6× lens. Your hands must be clearly visible. | REQUIRED |
| Recording mode | Normal speed. No slow motion, no time-lapse. | REJECTED IF ALTERED |
| Clip length | 2 min 45 sec minimum · most bounties run 5–15 minutes | REQUIRED |
| Claim code note | Handwrite your claim code + firstlylabs.com on a note. Place it on your work surface BEFORE filming — don't hold it up to the camera (see 05) | REQUIRED |
Every claim comes with a short code — you'll see it the moment you claim, on the task page while you hold it, and on the upload page. It exists for one reason: footage containing your code was provably filmed after you claimed the bounty. Original work, verified automatically.
Every interaction should be captured whole — the AI learns from complete sequences, not fragments. Never cut mid-task.
Pick the object up fully in frame
Do the action with it, visible throughout
Set it down without leaving frame
Let the result sit in frame a beat
We buy reality. Any post-recording modification is automatic rejection.
These are the automatic fails. But here's our difference: every rejection comes with the exact measurement that caused it — and you can always re-film and resubmit. A rejection costs you the time, never the account.
| Error | Description |
|---|---|
| Face visible | Any face — yours, a bystander's, a reflection, a photo on the wall |
| Handheld / propped | Phone held in hand or resting on a surface instead of mounted |
| Vertical video | Portrait orientation, any duration |
| Below 1080p30 | Resolution or frame rate under the floor |
| Edited | Any cut, filter, effect, or post-recording change |
| Slow motion / time-lapse | Any speed alteration |
| Too dark / blown out | Hands not clearly visible due to lighting |
| Idle footage | Long pauses — stay in motion doing the task |
| Personal documents | IDs, screens with personal data, mail, account numbers in frame |
| Not your footage | Stock video, re-records, AI-generated content — account closure, not just rejection |
No invoices, no monthly payroll calendar, no minimum wait beyond your tier. Accepted footage accrues to your balance and pays out automatically.
Accepted clips pay $3.75–$5.00 per footage-hour for household bounties (your measured capture tier sets the rate) and up to $25.00/hr for certified trades. Balance pays out in USDC on Solana once it clears $5.00, on your badge/holder schedule — as fast as 48 hours. Every payout carries your clip receipts.
Wallet: any Solana wallet (Phantom, Solflare, Backpack) or an exchange deposit address. Full rates, tiers & schedules on the main page.
Read it once, film it right, get paid. Your first bounty is waiting.
Complete your application →