8 Replies Latest reply on Nov 4, 2010 3:12 PM by atr_23

    Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?

    sfjedi Level 1

      I'm thinking of a tool where you could:

      1. Take an existing photo that has some kind of fill/gradient in it.
      2. Define what type of fill you have identified (e.g. radial, linear, etc.).
      3. Mask out regions of the photo that have nothing to do with the fill.
      4. Define the starting and ending points of the fill.

       

      For example, take this photo:

       

      dexter.jpg

       

      I've identified a radial fill:

       

      fill_type.png


      Now I want to mask out the parts of the photo that have nothing to do with the actual fill, as well as define where I think the starting and ending points should be. I've also excluded the JPG artifacts from the border of the gradient, because they had some darker colors in there that I don't want to be calculated. Note that the starting point in this example has no possible color that could be identified, but it may be *guessed* what the color should be, based on the rest of the fill, programmatically.

       

      span.png

       

      Lastly, I should be given the original fill when I complete the tool. This fill would have a starting and ending point, but also 3 intermediate colors defined in various locations. Also, the radial fill is up 15% on the Y axis. Programmatically, from the information provided above, this should be able to be calculated.

       

      fountain_fill.png

       

      Note. I'm using Corel PHOTO-PAINT to illustrate what I'm talking about here, because that's what I'm most comfortable using.

       

      Don't hate.

       

      Edit: In hindsight, I really don't think you should need to draw a line or specify which gradient you've identified. Ideally, that too should be handled automatically! Programmatically, this shouldn't be an issue. As a programmer, knowing how things work from the back end, I feel I'm licensed to say such a statement. Additionally, there should be a threshold slider that compensates for JPG compression and such by blurring the gradient a bit and getting rid of nodes that are too close together.

        • 1. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
          Reynolds (Mark) Level 4

          Hmm, you may not be aware but something ALMOST what you are looking for already exists in Photoshop - within gradient layers

           

          Picture 10.png

           

           

          At the moment its possible to have gradients definable in terms of their direction and size, in this way already. What you may be asking for here I suppose, is some kind of extension to the Gradient Layers capability? With definable vector points?This I think for illustrators would be a very useful addition.

           

          The MASKING side of what you have asked for is available already in about 50 different ways.

           

          You can also of course use Illustrator Smart Objects - this will allow Gradient Meshes, and the new gradient capabilites in Illustrator - all indirectly within Photoshop.

           

          At the moment the way Gradients are handled in Photoshop is very old, requiring clicking through embedded dialogs. In addition making contact and moving gradient points, and midpoints, at least on mac has become very difficult - so its all certainly in need of revision.

          • 2. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
            sfjedi Level 1

            I understand that the current gradient tool allows you to create gradients that are virtually the same, but I'm trying to say that *reproducing* gradients that you see in flattened images like JPGs or whatever is pretty damn near impossible; thus, there should be a tool that helps assist in reverse-engineering these gradients automatically so that you can reuse the same gradient with consistent results! This would be useful, especially, to clean up any JPG compression artifacts you might see as well.

            • 3. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
              Reynolds (Mark) Level 4
              but I'm trying to say that *reproducing* gradients that you see in flattened images like JPGs or whatever is pretty damn near impossible;

              Wouldn't go that far. You are probably not aware of this, but you just open the Gradient Editor dialog and just select the color points on the gradient with a single click- pulling those values with your eyedropper from the source image. Its pretty easy to recreate any standard gradient.

              • 4. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
                sfjedi Level 1

                Reynolds (Mark) wrote:

                 

                Wouldn't go that far. You are probably not aware of this, but you just open the Gradient Editor dialog and just select the color points on the gradient with a single click- pulling those values with your eyedropper from the source image. Its pretty easy to recreate any standard gradient.

                 

                Except, that's *exactly* what I tried to do in this example and could just never really seem to reproduce the "exact same" gradient. It was always off somehow. Also, the point here is to make an efficient reproduction of the existing gradient that you see. That is, instead of 10 or 20 intermediate colors in the gradient, maybe you only need 2-3? The tool should be able to figure out the lowest common denominator. And, like I said before, maybe there should be some kind of threshold slider if you think it needs more accuracy.

                 

                It would just make the whole process way more easy!

                • 5. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
                  Gyno-jiz Level 5

                  Jed:

                   

                  Re-read my 10-point post in your other thread. Slice out the grad, set the rulers to % and you can place points pretty much exactly. The info palette will tell you where to place them. I can tell you from experience you do not want to place 10 to 20 points on a grad, unless it's running the full gamut.

                   

                  I found that in your example, the manual process took me about 20 seconds. It took me about 10 seconds to cut the mask (for helping the computer out). And to quote you, something else I know from bitter experience: cutting masks is tedious. I'm definitely not saying this is a useless tool. But I wouldn't wish the PS engineers to stop working on more useful features.

                   

                  Why are you fixated on exactly the same grad? Close enough will be close enough. If you can't tell the difference with the images side by side, isn't that a match?

                   

                  How many of these do you envision doing (at 10-seconds saving a pop)?

                   

                  And why don't you have access to the original layer files with grad fills to drag to your new doc?

                   

                  My stance is that it is an automation feature only, and will do the user no better for quality than a manual process. But guessing the missing color behind the noggin, I can see some use. Would that be something a user would quickly learn? My first-shot guess in your experiment was dead-on (if I do say so myself).

                   

                  I've certainly duped grads before, but I think the thing I'm failing to understand is under which circumstances would this tool save anyone more than a few minutes a week. I'm sure it's just a limit to my imagination or experience...

                  • 6. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
                    sfjedi Level 1

                    J Maloney wrote:

                     

                    Re-read my 10-point post in your other thread. Slice out the grad, set the rulers to % and you can place points pretty much exactly. The info palette will tell you where to place them. I can tell you from experience you do not want to place 10 to 20 points on a grad, unless it's running the full gamut.

                     

                    I completely understand the step process of what you are suggesting, but it doesn't solve every scenario.

                     

                    I found that in your example, the manual process took me about 20 seconds. It took me about 10 seconds to cut the mask (for helping the computer out). And to quote you, something else I know from bitter experience: cutting masks is tedious. I'm definitely not saying this is a useless tool. But I wouldn't wish the PS engineers to stop working on more useful features.

                     

                    Suggesting a feature does not in any way suggest that PS engineers should "stop working on more useful features." My feature requests are simply that–feature requests! Whether great or small, they still remain feature requests. However Adobe decides to prioritize them (if they decide to implement them at all) is a completely different story.

                     

                    And cutting masks? Yeah, if you need to deal with feathering and all that it can be quite tedious, but this is just a simple mask around the edges of the baby.

                     

                    Why are you fixated on exactly the same grad? Close enough will be close enough. If you can't tell the difference with the images side by side, isn't that a match?

                     

                    If you can't tell the difference, then yes! Problem is–I've found–sometimes it takes a bit of trial and error to get it just right and I'd rather do it right the first time!

                     

                    How many of these do you envision doing (at 10-seconds saving a pop)?

                     

                    Not many. Again, refer to the previous remark about Adobe's choice of priority. I'm not suggesting this is SO IMPORTANT OMG PLZ ADD THIS FEATURE NOW!!!

                     

                    And why don't you have access to the original layer files with grad fills to drag to your new doc?

                     

                    Because, in this case, it was an online, Flash-based tool that was having server issues and not responding! Instead of using their crappy tool with horrible compositing, I decided to do it manually to make it look *way* better than the product they were going to produce. Of course, I wouldn't have access to some online, Flash-based tool's image. It's not like they had a PSD download or something. It wasn't mine in the first place! I had to screenshot it just to get it into Photoshop!

                     

                    I've certainly duped grads before, but I think the thing I'm failing to understand is under which circumstances would this tool save anyone more than a few minutes a week. I'm sure it's just a limit to my imagination or experience...

                     

                    Your confusion is all aimed at priority, once again. In my opinion, even if it saves one minute a week, that's enough to add the feature. I'm all about efficiency, shaving off minutes or seconds anywhere they can be shaved off! Of course, tackle the larger efficiency obstacles first. Doing anything else would be–well, inefficient!

                    • 8. Re: Reverse-engineer Gradient Tool or Gradient Eye-dropper Tool?
                      atr_23

                      I want this feature too. Microsoft Expression Blend has it too.