SightAPIOCR v1
Video OCR API · ready

Video OCR that makes every word usable.

SightAPI turns on-screen text into structured JSON with timestamps, confidence, and bounding boxes—so your team can search, moderate, index, and automate video.

Process your first video Read the API docs
No card required1,000 monthly creditsFailed jobs cost 0
ocr-result.json completed
{
  "text": "ORDER #8492 CONFIRMED",
  "confidence": 0.94,
  "start_ms": 4250,
  "end_ms": 5790,
  "bounding_box": {
    "x": 0.1875,
    "y": 0.2222,
    "width": 0.5156,
    "height": 0.1667
  }
}
4.25sstart
94%confidence
1.54sduration

Normalized coordinates · source timing preserved

Built for real workflows

A video-text layer your product can trust.

One focused API replaces a fragile stack of media decoding, OCR models, temporal tracking, retries, and output storage.

01

Structured output

Every detection includes text, exact timestamps, confidence, and normalized bounding boxes—not a flattened transcript.

02

Production controls

Idempotency keys prevent duplicate work. Job stages, retries, errors, and final charges remain traceable.

03

Managed execution

Decoding, change-aware frame selection, OCR inference, tracking, and rendering run without GPU operations on your team.

Transparent usage

Pay for measured processing, not every source frame.

Change-aware selection and perceptual frame hashing avoid unnecessary OCR work. Each completed job reports processing time and charged credits.

See transparent pricing
2 creditsper measured GPU-worker second
0 creditsfor failed jobs
Nativesource timing and frame rate
Includedtracking and structured result storage

Three-minute path

From local MP4 to structured OCR.

Create an API key, upload a video, and start an idempotent job.

Full quickstart Playground
cURL quickstartbash
# Create an upload target
ASSET=$(curl -s -X POST "$SIGHT_API_URL/api/v1/assets" \
  -H "Authorization: Bearer $SIGHT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"filename":"sample.mp4","content_type":"video/mp4"}')

# Upload, then start an idempotent OCR job
curl -X PUT "$(echo "$ASSET" | jq -r .upload_url)" --upload-file sample.mp4
curl -X POST "$SIGHT_API_URL/api/v1/jobs" \
  -H "Authorization: Bearer $SIGHT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input_asset_id":"asset_id","operation":"OCR"}'

Evidence over claims

Quality metrics that stay inspectable.

Our reproducible benchmark reports recognition accuracy, spatial quality, short-lived text recall, processing time, and cost.

Character & word error rate
Bounding-box IoU
Text visible <1 second
Time and cost per video hour
Duplicate detections
Dataset and test scripts

Start with 1,000 credits

Turn video text into product data today.

Create a free account