OCR
Recognize pasted images once or monitor a cropped shared-screen region live. Configure Tesseract, record results, compare text, and trigger alarms.
Clipboard
Paste an image or upload a JPEG, PNG, SVG, or WebP file. OCR starts automatically when a new image is added.
Waiting for an image.
Share Screen
Share a screen or window, then select a cropped region for OCR. Recognition runs every second after a region is selected.
Captured Region
Click and drag a crop region on the shared screen, or enter Left / Top / Right / Bottom coordinates.
OCR Text
Configure Tesseract.js below. Images receive a 3x upscale pass before recognition.
Tesseract.js will load on first OCR run.
Alarm and Notification
Bright Chime / 100% / Sound Off / Notifications Off
Compare and Decide
Compare normalized OCR lines with the pool and target groups. Leave Pool empty to compare directly with target lines.
| Line | OCR | Picked | Similarity | Accepted |
|---|
Record
Record each completed Live OCR result with elapsed time.
Recording is stopped.
Tesseract Reference
This tool uses Tesseract OCR through the browser-based Tesseract.js WebAssembly wrapper.
For faster, more accurate recognition, select a narrow cropped region focused closely on the words instead of running OCR on the entire screen.
| Input | Recommended | Why use this selection |
|---|---|---|
| Language | eng | Uses the English traineddata for English text. |
| OEM | 3 | Lets Tesseract select its available default recognition engine. |
| PSM | 4 | Works well for a narrow cropped region containing one column of text with variable-sized lines. |
| Preserve Spaces | 1 | Retains relative spacing between recognized words. |
| Code | Language | Usage |
|---|---|---|
eng | English | Default for English text. |
spa | Spanish | Spanish traineddata. |
fra | French | French traineddata. |
deu | German | German traineddata. |
ita | Italian | Italian traineddata. |
por | Portuguese | Portuguese traineddata. |
chi_sim | Simplified Chinese | Simplified Chinese traineddata. |
chi_tra | Traditional Chinese | Traditional Chinese traineddata. |
jpn | Japanese | Japanese traineddata. |
kor | Korean | Korean traineddata. |
eng+spa | Multiple languages | Join installed language codes with +. Commas and spaces entered above are converted to this format. |
| Value | Engine | Behavior |
|---|---|---|
| 0 | Legacy | Uses only the legacy Tesseract recognition engine and requires compatible legacy traineddata. |
| 1 | LSTM | Uses only the neural-network LSTM recognition engine. |
| 2 | Legacy + LSTM | Combines both engines and requires compatible legacy traineddata. |
| 3 | Default | Lets Tesseract choose the available engine. |
| Value | Mode | Expected layout |
|---|---|---|
| 0 | OSD only | Orientation and script detection only. |
| 1 | Auto + OSD | Automatic page segmentation with orientation and script detection. |
| 2 | Auto, no OCR | Automatic page segmentation without OSD or recognition. |
| 3 | Auto | Fully automatic page segmentation without OSD. |
| 4 | Single column | One column of text with variable-sized lines. |
| 5 | Vertical block | One uniform block of vertically aligned text. |
| 6 | Single block | One uniform block of text. |
| 7 | Single line | One text line. |
| 8 | Single word | One word. |
| 9 | Circle word | One word arranged in a circle. |
| 10 | Single character | One character. |
| 11 | Sparse text | Find as much text as possible without assuming an order. |
| 12 | Sparse text + OSD | Sparse text with orientation and script detection. |
| 13 | Raw line | One text line, bypassing Tesseract-specific line-layout heuristics. |
| Value | Mode | Behavior |
|---|---|---|
| 0 | Normalize spaces | Uses normal word separation without retaining proportional gaps from the source. |
| 1 | Preserve spaces | Retains relative spacing between recognized words; this page's recommended default. |