#Journal Recognizer: OCR for handwriting

1 messages · Page 1 of 1 (latest)

fathom pond
#

https://chat.openai.com/g/g-T7bW2qVzx-journal-recognizer

The app, named Journal Recognizer, is designed to transcribe handwritten notes and present them in a markdown code block format. It's perfect for anyone looking to digitize handwritten journals, notes, or scribbles! It includes descriptions for non-text elements, and a summary of the content.

Multiple images in the same prompt will be treated as a single document, and output one block of text.

A prompt with only an image triggers the OCR with default settings. Enter further special instructions (ex: add hard returns to preserve the words in each line) in the prompt with the image or images. Images should be loaded in the order of the handwritten pages.

This is alpha test. Looking for some feedback.

plush rose
#

Does this use external api?

fathom pond
#

No, just ChatGPT.

#

I don't know the limit on the number of images in a single prompt. It took 8 pages no problem.

kind zodiac
#

working perfect

fathom pond
#

Ah, 10 file limit (ChatGPT limit)

kind zodiac
#

it can read the unreadabe scribbles. How can this be

fathom pond
# kind zodiac it can read the unreadabe scribbles. How can this be

I've been fine-tuning all day, trying out very old, even pretty hard to read old image files. It's not at all perfect, but it's smart enough to use the context of the sentence to tease out the words. At least that's what I think it's doing. I've wanted this for my notebooks for about 40 years. I've tried and gave up on OCR for years. This is not only usable, it's snappy!

#

I mistakenly added Code Interpreter for a few minutes, which really screwed things up. If you got some weird results a little while ago, try again.

kind zodiac
#

Yes true. you cant seesee what I write to gpts. Am ı right? I will use your gpt allall the time . Please never delete

fathom pond
#

Thank you! I'm thrilled it's useful for you. Have any requests for features? Have you tried pages that include drawings? Are you using this on the computer or mobile?

kind zodiac
plush rose
#

Im trying to add this to another thing, what are the instructions?

pearl igloo
plush rose
pearl igloo
#

"I want to add your instructions to my GPT" isn't really sharing knowledge

plush rose
#

You maybe cite the whole question

#

@pearl igloo Your insights are truly unique, but I must turn my attention elsewhere. Perhaps you can find someone else to appreciate them?

sand copper
#

Something is broken. Tried two images, failed both times. Unfortunately, the transcription process is still encountering technical issues, and I'm unable to transcribe the handwritten text from the latest image you provided.

fathom pond
#

The OpenAI's systems had downtime last night here and there. Try again. As for instructions... you ask it to do things, then it tries this and that, finally does what it wants.

fathom pond
#

New improvements, faster, more efficient. Can swallow 10 images and spit back one paragraph delimited text with relevant graphic descriptions in the capture plaintext copy codeblock.

kind zodiac
#

hi. Summarization can be an individual prompt. Ocr alone is much better. It can give only the text

jovial harbor
pearl igloo
# fathom pond

ChatGPT just did the same. How is it different considering custom GPT's take longer than plain ChatGPT?

fathom pond
#

Try doing ten pages at once. I want to be able to scan 100 pages in 10 minutes or so, and just paste them into the document I'm making. I don't want to copy one page at a time. One click gets 10 pages. And you can specify special handling, but it's optimized for lots of pages with few clicks, and little text processing to get rid of hard returns in the middle of paragraphs, the bane of every OCR project I've embarked on.

fathom pond
pearl igloo