[HN Gopher] Stable Doodle: Transform your doodles into real imag...
       ___________________________________________________________________
        
       Stable Doodle: Transform your doodles into real images in seconds
        
       Author : DerekBickerton
       Score  : 36 points
       Date   : 2023-07-19 19:04 UTC (3 hours ago)
        
 (HTM) web link (clipdrop.co)
 (TXT) w3m dump (clipdrop.co)
        
       | og_kalu wrote:
       | It's the T2I adapter rather than control
       | net(https://github.com/TencentARC/T2I-Adapter). Other than that,
       | the big thing is that it uses SD XL
        
       | tudorw wrote:
       | Nice, I'm a bit confused by the existence of clipdrop api and
       | stability api, will they converge?
        
         | jeron wrote:
         | from what I understand, clipdrop api offers these features[0]
         | while stability api is mainly stable diffusion which is image
         | generation from prompting[1] (if you are referring to stable
         | diff 2.1)
         | 
         | these APIs provide pretty different functionalities
         | 
         | [0]:https://clipdrop.co/apis
         | [1]:https://huggingface.co/stabilityai/stable-diffusion-2-1
        
       | simbolit wrote:
       | it doesn't work. it just creates images based on the prompt, it
       | has nothing to do with the doodle at all.
       | 
       | Edit: after some trying, there is sometimes some rough shape
       | matching, but it's often very much off.
        
         | boredemployee wrote:
         | agree, image has nothing to do with the doodle
        
           | myhf wrote:
           | try drawing a different shape input, the output shape will
           | match it
        
           | ClassyJacket wrote:
           | It does, but not reliably:
           | 
           | https://imgur.com/a/LqxBWlA
        
         | WheatMillington wrote:
         | Yep, I don't think it used my doodle at all for the generation.
        
         | simbolit wrote:
         | try doodling a cat and prompting 'bird' or vice versa and you
         | see what I mean.
        
           | bb88 wrote:
           | So it doesn't try to interpret the doodle. But as said
           | elsewhere the output shape will try to match it.
        
             | nashashmi wrote:
             | I found the same thing. Drew a lousy horse looking at it
             | from the side and it gave me horse from the front.
        
               | schoen wrote:
               | I drew a man holding a sword and got men holding swords
               | in completely different postures.
               | 
               | It probably was trying to match some elements of the
               | shapes, but it certainly didn't match an ordinary human
               | understanding of what the items in my drawing were
               | supposed to represent!
        
       | WheatMillington wrote:
       | After 3 attempts in just starting to understand it before being
       | locked out. I'm 100% not starting an account to keep playing with
       | this.
        
       | rootusrootus wrote:
       | Eh, I don't think this is better than
       | https://scribblediffusion.com/. None of the three images were at
       | all like my drawing, but Scribble Diffusion did include the right
       | elements in the right places.
        
         | simbolit wrote:
         | Can confirm, this (scribblediffusion) is much better!! I
         | doodled a cat, and then prompted it with "cat", "bird", and
         | "snake". All three look like my doodle, the cat looks great,
         | the other two look forced, just as one would expect.
         | 
         | With stabledoodle, it gives me a cat, then a bird, and then a
         | snake, and I can't find my doodle in any of them.
        
       ___________________________________________________________________
       (page generated 2023-07-19 23:00 UTC)