This update includes significant updates to capturing text from images (screenshots) and audio/video files.
Audio Transcription Timestamps
When transcribing text from audio or video files, TextBuddy can now (optionally) provide the timestamps associated with the words spoken.
You can view the raw timestamps of each detected word, or you can let TextBuddy group phrases into more manageable chunks of time. Here’s a demo:
Visual OCR Results
The other major improvement in this update is TextBuddy uses a new visual algorithm when extracting text from images and screenshots. Here’s a contrived example and then a few demo videos.
Say you use TextBuddy to OCR the portion of this webpage outlined in red.
Previously, you’d get results like this:
Yes, the text is recognized accurately, but because of the two column layout, the words from both news articles are incorrectly intermingled.
With today’s TextBuddy update, the app attempts to visually format the results as they appear on screen like this:
Here’s a live demo taken from a scanned PDF:
Of course, there are complexities when it comes to trying to position plain text that can’t overlap in a document based on images which do allow overlapping text and other oddities such as multiple fonts, sizes, and font weights. To help with this, TextBuddy allows you to visually adjust the spacing between words with a slider to better match the results you’re looking for.
Here’s an example of me grabbing source code from an Apple developer video presentation.
First, the old way:
And the new way showing off the adjustment slider:
Visual formatting does the best it can, but at times it could absolutely be more problematic than helpful – especially if you are just grabbing a simple text layout. It can be disabled on a case-by-case basis.
I hope you find these improvements to audio transcription and OCR helpful. This TextBuddy update also includes the changes below.
- New: You can choose a default language in Preferences that will be used when recognizing text in images.
- New: Added “Keep Only Dates” command, which parses out and returns any date-like strings it can find.
- New: Added “Keep Only Phone Numbers” command, which parses out and returns any phone numbers it can find.
- New: Added “Keep Only Addresses” command, which parses out and returns any address-like strings it can find.
- New: Show a progress spinner during long-running commands so you know TextBuddy is still doing something.
- Improved: The text editor will lock during long-running commands to prevent accidental edits while processing is happening.
- Improved: Word wrapping gigantic documents should no longer freeze the app. (It may still take a while to finish.)
- Improved: “Word Wrap and Respect Code Comments” should be much faster now.
- Improved: Better error messaging when transcribing audio fails.