Live data from Hacker News

Realistic computer-generated handwriting

calligrapher.ai

51–60 of 247 posts

Re: Realistic computer-generated handwriting

#52

Should it worry me that the legibility slider cranked all the way to zero is still more legible than my actual handwriting?

I always wondered if pharmacists had any specific training in the terrible handwriting doctors put on prescriptions, or if they all just figure it out in their first year.

Re: Realistic computer-generated handwriting

#56
post #23

Awesome, but fails with accents etc. A case for polish: „Zażółć gęślą jaźń” - fails in many gunny ways. I also tried emojis - of course it failed, but now I’m super curious how would hand written emojis look like :D

It would probably look a lot like handwritten kanji.

Asking various stable diffusion implementations, it looks like it agrees with me at least a little. When anything close to what I want is generated it tends to actually contain kanji like fire and life.

Re: Realistic computer-generated handwriting

#57

Should it worry me that the legibility slider cranked all the way to zero is still more legible than my actual handwriting?

For accuracy, it needs negative legibility settings. Anyone who's been in a sprint retro with hand written action points knows this.

Re: Realistic computer-generated handwriting

#58
post #10

That's not really realistic, it can't even do normal cursive handwriting with first capital letter joined, I dunno anyone who writes like any of those 10 styles, maybe they are common in US, but certainly not in this part of Europe. Get back to me when it will knows at least something like this: https://qph.cf2.quoracdn.net/main-qimg-5d1b0e1d28c12813e1255... or here especially 1st and 3rd line https://d50-a.sdn.cz/d_…

Realistic for me would be including small ink pools, skips, smudges, crossed out corrections, and so on.

Re: Realistic computer-generated handwriting

#60
My first name contains one of the fewest used letters of the alphabet and in first position nonetheless: Xavier

Let me tell you: the training data did not contain that many X to begin with :D

The Xs at best look like Ts, at worst like malformed Os.

Always funny to find the blind spots of some piece of tech.

Post reply on HN