Use Cases
From user-generated content to LLM training data, FirstHandAPI powers auto-annotated, human-captured data collection with one API call. You set the price per file; 20% platform fee; you pay only for 3+ star files. GPS-verified capture when your job requires it. mcp-native: npx @firsthandapi/mcp-server
User-Generated Content (UGC)
Collect authentic photos, videos, and audio captured by real people on their own phones. Set require_location on a job and uploads must carry device GPS — captures outside your target radius are rejected before you ever see them. Every approved file comes with auto-generated annotations and area-level capture coordinates.
- Storefront and venue photos with signs auto-extracted via OCR
- Product-in-use lifestyle photos from real homes, not studios
- Audio reviews and voice testimonials from real customers
- Short video walkthroughs with scene and action labels included
curl -X POST https://api.firsthandapi.com/v1/jobs \
-H "Authorization: Bearer fh_live_..." \
-H "Idempotency-Key: ugc-campaign-001" \
-H "Content-Type: application/json" \
-d '{
"type": "data_collection",
"description": "Photograph the storefront of any independent coffee shop. Include the sign and entrance in natural lighting.",
"files_needed": 50,
"accepted_formats": ["image/jpeg", "image/png"],
"price_per_file_cents": 200,
"require_location": true,
"location": {
"city": "New York",
"state": "NY",
"latitude": 40.7128,
"longitude": -73.9860,
"radius_km": 10
}
}'Ground Truth & Evaluation
Build verified reference datasets with structured annotations included on every file. Object labels, OCR text, scene classifications, and transcripts give your ML team ready-made ground truth for evaluating model accuracy and detecting regressions.
- Street signage photos with auto-extracted OCR text
- Ambient and speech audio with speaker counts and timestamped transcripts
- Real-world video with scene descriptions and action labels
- Multi-modal evaluation sets combining images, audio, and video
curl -X POST https://api.firsthandapi.com/v1/jobs \
-H "Authorization: Bearer fh_live_..." \
-H "Idempotency-Key: eval-set-v3" \
-H "Content-Type: application/json" \
-d '{
"type": "data_collection",
"description": "Photograph street signs clearly, one sign per image. Must be legible and in focus.",
"files_needed": 200,
"require_location": true,
"accepted_formats": ["image/jpeg", "image/png"],
"price_per_file_cents": 150,
"location": {
"city": "New York",
"state": "NY",
"latitude": 40.7580,
"longitude": -73.9855,
"radius_km": 5
}
}'LLM Training Data
Source diverse, geo-tagged media at scale for multimodal fine-tuning. Every file arrives with structured annotations — object labels, scene classification, transcripts, color palettes — so your training pipeline starts with rich metadata, not raw blobs.
- Street-level photos with auto-generated object and scene labels
- Accent-diverse speech recordings with timestamped transcripts
- Short-form video clips with scene and action tracking
- Multi-language audio from contributors on 5 continents
curl -X POST https://api.firsthandapi.com/v1/jobs \
-H "Authorization: Bearer fh_live_..." \
-H "Idempotency-Key: training-batch-042" \
-H "Content-Type: application/json" \
-d '{
"type": "data_collection",
"description": "Record 30-60s of natural conversation on any topic. Clear audio, minimal background noise.",
"files_needed": 250,
"accepted_formats": ["audio/mp4", "audio/mpeg"],
"price_per_file_cents": 300,
"location": {
"city": "New York",
"state": "NY",
"latitude": 40.7282,
"longitude": -73.7949,
"radius_km": 15
}
}'Screen Capture & Recording
Collect app screenshots, workflow recordings, and UI testing data from real devices. Ideal for QA teams validating cross-device rendering, UX researchers studying real user flows, and teams building UI understanding models.
- Cross-device screenshot testing for responsive design
- User workflow recordings for UX research
- App store screenshot generation across locales
- UI interaction data for accessibility auditing
curl -X POST https://api.firsthandapi.com/v1/jobs \
-H "Authorization: Bearer fh_live_..." \
-H "Idempotency-Key: screen-capture-sprint-7" \
-H "Content-Type: application/json" \
-d '{
"type": "data_collection",
"description": "Record a 60s walkthrough of the onboarding flow. Show each screen clearly.",
"files_needed": 100,
"accepted_formats": ["video/mp4", "video/quicktime"],
"price_per_file_cents": 500
}'What you get back
Every approved file arrives with structured annotations attached — this is a real response from a production job, field names unchanged. Images carry objects, OCR, scene, colors, and safety scores. Audio adds timestamped transcripts, speaker counts, and SNR. Video adds scene descriptions, action labels, object tracking, and keyframes.
- Only 3+ star files are delivered — you never pay for rejects
- Estimated confidences on every label (model-estimated, not calibrated — we say so)
- Safety screen on every file: NSFW, violence, visible PII
- Private delivery: org-scoped API, expiring signed URLs
{
"type": "image",
"objects": [
{ "label": "utility pole with transformer and cables",
"confidence": 0.94, "position": "center-right" },
{ "label": "parked cars", "confidence": 0.85 },
{ "label": "pedestrians with bicycles", "confidence": 0.82 },
{ "label": "concrete building", "confidence": 0.86 }
],
"scene": {
"setting": "interior of a bus looking out onto a suburban street",
"indoor": true, "confidence": 0.9
},
"text_extraction": { "full_text": "CAIO TECGLASS GNV" },
"color_palette": ["#7FA8CB", "#1B2733", "#9EA7A5"],
"face_count": 2,
"quality_metrics": {
"blur_score": 0.18, "exposure": "normal", "noise_level": "low"
},
"safety": {
"nsfw_score": 0, "violence_score": 0, "pii_detected": false
},
"annotation_model": "claude-opus-5"
}Ready to collect real-world data?
Sign up in under two minutes, get $2.50 in free credits, and post your first data collection job today.