SynthOCR

Make OCR training data

Choose your text and a look. You get text images, each with its exact text label (ground truth), saved to a folder on your computer.

1Text

Languages
Where the text comes from

2Look

Image style
Customize
75%

The rest are clean black text on white.

Backgrounds

Each image with effects uses one of the backgrounds you select.

Effects

Auto uses each background's own amount.

Tilt, blur and contrast

Each image gets a random amount up to the value you set. 0 turns it off.

2°
0.55 px
10%
48 px

How tall the letters are in pixels. Long lines shrink to fit.

Fonts All

Each image uses one of the chosen fonts that can draw all of its text. Fonts you add stay in this app on your computer.

Preview

Rendering examples…

Text label (ground truth)

…

3Generate

Save to
No folder chosen

Choose a folder on your computer. The dataset is saved there when it is ready.

A new folder with this name is created. Letters, numbers, - and _.

Label format