OCR

OCR extracts text from images, scans, and document screenshots.

After recognition, you can copy the result, export it as Markdown, PDF, or Word, or package multiple formats together for download.

What OCR Can Do

Feature	Description
Image text recognition	Extracts text from images, screenshots, and scans.
Document layout recognition	Better for tables, formulas, stamps, and mixed text-image layouts.
Multiple services	Supports Baidu PaddleOCR, Microsoft Azure Vision, and Google Vision.
Copy results	Copy recognized text after processing.
Export files	Export Markdown, PDF, and Word.
Batch packaging	After recognizing multiple files, download results as a package.

Configure OCR Services First

Open:

System Settings -> Other Settings -> OCR

Fill in credentials for the services you want to use:

Service	What To Enter	Best For
Baidu PaddleOCR	PaddleOCR Token	Recommended first choice. Good for documents, images, tables, and mixed layouts.
Microsoft Azure Vision	Azure Vision Endpoint and Azure Vision API Key	Useful if you already use Microsoft cloud services.
Google Vision	Google Vision API Key. Service account JSON is only used for quota query.	Useful if you use Google Cloud services.

Save after filling in credentials.

You can configure only one service for initial testing. You do not need all three.

Google Vision Setup

Google setup has two parts:

Goal	Requirement
Use OCR	Enable Cloud Vision API, then create an API Key.
Query usage	Create a service account, grant Monitoring Viewer, then download the service account JSON.

Use Google for OCR

1. Open Google Cloud Console.
2. Go to APIs & Services.
3. Open Library, search for Cloud Vision API, and enable it.
4. Return to Credentials.
5. Create an API Key.
6. Open the API Key and copy it.
7. Paste it into Google Vision API Key in ImgBed.
8. Save.

You can then choose Google Vision in the OCR dialog.

Query Google Usage

Quota query is not required for recognition.

It only shows roughly how many Google Vision calls were used in the last 30 days.

1. In Google Cloud Console, open IAM & Admin.
2. Open Service Accounts.
3. Create a service account, such as vision-monitor.
4. Grant it the Monitoring Viewer role.
5. Open the service account details and create a key.
6. Choose JSON.
7. Download the generated JSON file.
8. Return to ImgBed and import it under service account JSON (optional).
9. After import succeeds, click quota query.

After import, ImgBed shows the project name that owns the service account. When querying usage, ImgBed reads Google monitoring data and shows this month's call count.

In short:

Item	Purpose
Google Vision API Key	Performs OCR recognition.
Service account JSON	Queries how many Google Vision calls were used.
Monitoring Viewer role	Allows the service account to read usage data.

Get a Baidu PaddleOCR Token

Baidu PaddleOCR requires an access token.

Open the API call window on the Baidu PaddleOCR page, click to get a token, then copy it.

Return to ImgBed, paste it into PaddleOCR Token, and save.

Start Recognition

In File Management, select an image or document screenshot and click OCR.

In the dialog, choose the recognition service and model.

Common PaddleOCR model choices:

Model	Best For
PP-StructureV3	Recommended default. Good for documents, tables, formulas, stamps, and mixed layouts.
PP-OCRv5	Simple images, ordinary text, and lightweight recognition.
PaddleOCR-VL	Multilingual, complex images, and chart-like content.
PaddleOCR-VL-1.5	More complex document pages and layout recovery.

If you are unsure, start with PP-StructureV3.

Advanced Options

Option	Description
Orientation correction	Use when the image is rotated or skewed.
Document flattening	Use for photographed documents with curvature or tilt.
Layout detection	Use when you want to preserve headings, paragraphs, tables, and image structure.
Chart recognition	Use when the image contains charts or complex structures.
Beautify Markdown	Makes exported Markdown easier to read.

For regular screenshots, keep options minimal. For document scans, enable more document-related options.

View Results

After recognition finishes, the dialog shows the result.

You can copy it directly or choose export formats.

For document pages, exported PDF can preserve page appearance while keeping text searchable. This is useful for archiving scans and finding content later.

Choosing an Export Format

Format	Best For
Markdown (.md)	Notes, documentation systems, and later editing.
PDF (.pdf)	Preserving page appearance and scanned document results.
Word (.docx)	Continued layout editing, text modification, and handoff to others.
Export all	Saves multiple formats and the original image, suitable for important archives.

If you only need text, export Markdown.

If you need page appearance, use PDF or Word.

Word Output

Exported Word documents can be opened and edited with office software.

Some documents include recognized images, headings, and paragraphs in the Word output.

Recognition quality depends on original image clarity, model choice, and document complexity.

Best File Types for OCR

File Type	Recommendation
Clear screenshots	Recognize directly.
Scans	Prefer PP-StructureV3.
Photographed documents	Enable orientation correction and document flattening.
Tables, formulas, stamps	Prefer structured models.
Simple short text images	PP-OCRv5 is usually enough.

Clearer images with straighter text usually produce better results.

Common Cases

Case	Meaning
Recognition fails	Check that the service token or key has been saved.
Recognition is slow	Complex documents and large images take longer.
Table is incomplete	Try a structured model.
Text has mistakes	Blur, glare, and skew increase recognition errors. Try a clearer image.
Word output contains many images	Structured models may preserve some recognized images. This is normal.

Google Quota Query Fails

Check:

1. Service account JSON has been imported.
2. The service account has the Monitoring Viewer role.
3. Cloud Vision API is enabled for the project.

If you only need OCR and not usage query, you can ignore the service account JSON and only fill in Google Vision API Key.

Quick Flow

Open System Settings
-> Open Other Settings
-> Fill OCR service credentials
-> Save
-> Return to File Management
-> Select a file and click OCR
-> Choose a model
-> Wait for recognition
-> Copy results or export Markdown / PDF / Word