Support the GITP forums on Patreon
Help support GITP's forums (and ongoing server maintenance) via Patreon
Results 1 to 2 of 2
  1. - Top - End - #1
    Ettin in the Playground
     
    AsteriskAmp's Avatar

    Join Date
    Dec 2009
    Gender
    Male

    Default The Flower's Descent - A Big Art Project Post-Mortem

    About 2 months of ago I decided to finally draw something that took more than a single seating. It resulted in quite an enjoyable revision of topics which normally don't come into bearing for shorter projects like colour management and flow planning. It also involved nightmarish scenarios with my software of choice, design decisions and colour management. This is an insufficiently illustrated journey through the experience. Software of choice was Adobe Fireworks CS6 (this will be relevant in the worst ways possible) and initially keyboard and mouse but then because of carpal tunnel I had to resort to using a Wacom Intuos Tablet for what I was previously freehanding with the mouse.

    But first, the finished piece.
    Spoiler: Unless your monitor is 4k you won't get the full resolution and would be better served by clicking the above link or opening the image in a new tab
    Show



    A too long; didn't read because the post is rather long:
    • Doing the major contours thoroughly saves major time down the line avoiding costly replanning and rebalancing of colour, I did not do this.
    • Knowing the limitations of your software before you hit them can save you hours of pain down the line, I did not do this.
    • Colour Balancing and Distribution is super important and fun. Specially if done early, I did actually do this.
    • Big projects help bring into bearing techniques you'd normally wouldn't come into contact and force you to deal with far more constrained scenarios. It's much harder to fix lighting and perspective when it's 90% done and you just have to roll with it, this also means you learn limitation you might have been hiding from.
    • I need to practice drawing clouds in more scenarios and non-standard light sources.
    • I regret not taking the time to properly draw the background of the top section, making each tree by hand and properly drawing the mountains.
    • I should probably finish learning Illustrator and never again touch Fireworks.
    • Drawing in a tablet might alleviate Carpal Tunnel Syndrome.
    • Wood fiber can be approximated with extreme motion blur on rendered noise and posterization.


    Initial Plan

    The initial concept was considerably simpler and smaller. I sketched a general distribution of the canvas and where the focal elements would be and their thematic. From the beginning the plan was to do three oriental themed pieces interlinked by fluid motion across the canvas. The palette would hinge on red, white and black with the blues from the sky peering shyly from the clouds.

    The initial sketch was also VERY rough, the characters, the idea of connection through the water movement and the rough ideas for each section somewhat survived but in VERY different forms. This would become an issue constantly through the process for many reasons. On hindishgt, the physical medium technique of going over the entire thing with pencil would have saved me considerable amounts of stress and time.

    Spoiler: The Initial Planning Sketch
    Show

    Setting milestones and sectioning

    At this point sectioning is basically required to be able to get meaningful indicators of progress. In physical media sectioning normally comes naturally from z-position and the order in which you do sections is also forced by the fact that it's far easier to paint on top than painting around previously painted things, but since layers make this avoidable I sectioned off based on equal amounts of effort/time required.

    Adobe Fireworks (previously Macromedia Fireworks, which Adobe bought only because they wanted Flash and they got them as a package) is a 32-bit program. This might not seem relevant except it imposes a limit on how much RAM it can address, 4GB total. In practice it's around 3GB because of overheads and you should also keep clear from this because at saving time it might need additional memory. Adobe Fireworks is currently deprecated and has no 64-bit version unlike all of the Adobe CS Suite. If Fireworks runs out of memory, it will keep everything on screen but you cna't really dfo anything, including saving. This means the message out of memory is as good as a crash for almost every single case (occasionally you might get lucky and undoing the last action or hiding several big objects will let you save but there's no guarantee).

    To avoid running into this I sectioned off the sketch more aggressively than I'd otherwise do it, each section would be a file on its own, sections contained in other's would be flattened when working on those (reducing from hundreds of thousands of vertices and curves into a single bitmap). My sectioning proved more aggressive than needed and not aggressive enough. This was further compounded when the sketch was modified.

    Spoiler: The Sectioning of the Sketch
    Show

    Spoiler: The Sectioning that Ended up being used
    Show

    Starting Work

    Following the section plan the boat was the first section tackled. From the beginning one of my main concerns was having "running out of colours". Because I wanted each section to have a characteristic palette but also follow the main colour scheme of the whole piece I intentionally limited colour choice for the more detailed sections. With the power of the four colouring theorem and judicious breaking of the palette this wasn't a major limitation but at several point I was severely tempted to break it.

    The boat design was relatively quick. I started referencing from Renaissance paintings, a trend that would continue until becoming a design choice till the end. The wood fiber was tricked out from noise, extreme motion blur and finally posterization. The prow and the stern were manually done however, this required a day to do; not because it took long but because carpal tunnel flared up and by the end I couldn't use a mouse for the rest of the day.

    I was unhappy with the initial hair sketch, it didn't really look or feel oriental enough, the hair had several revisions of which only the first and last were documented at all.

    Spoiler: The First Day of Boat Work
    Show

    An intermediate step with the legs hanging out of the boat, while more visually interesting than the final result, didn't actually work with the rest of the piece tone wise.

    Spoiler: Legs Hanging out of the Boat
    Show

    I did a quick sample to have an idea of what the upper section would look like, mainly for the framing by the clouds and the moon+moonbow. I had done some work on the rocks and the tree and I completed the trunk and above quickly. This lead to the first major redesign. The lighting interaction with the tree was unsatisfactory and repeated tree redesigns still did not yield a satisfactory moon ray profile that would light up the boat and girl, this wasn't helped by the umbrella angle.

    I considered redoing the umbrella and restarting the design. The quick sketch was unlikely to scale well however, the background forests and mountains would completely change the framing and hte lighting was most likely going to end even more unsatisfactory when trying to deal with the perspective on the farther right edge and accounting for mist and obstructions.

    Spoiler: The First Upper Section Draft
    Show

    Technical Difficulties

    However, while moving the moon around a serendipitous discovery was had. Raising the upper canvas bound, moving the moon all the way up and letting clouds and mountains do the framing had great results in quick drafts.

    Using Photoshop Cloud Renderer for making flat fog planes and then using Fireworks perspective tool, vector masks for fitting it to the vector mountains and pushing Fireworks live effects for vector objects to produce the background layers the mountains and lake took shape. The starry sky was created using the noise generator and the screen layer mode. The clouds were a mix of Photoshop Cloud Renders, spherical distortions and manual vector fixes with carefully layered vector masks with colouring effects done in vector layers.

    This was the point were all hell broke lose. Fireworks reached the RAM limit REALLY fast. Not only that, it did so mid way through. Because of how the layers interacted I couldn't properly section this off into more parts, everything that could be split to their own file already was. My choices were moving over to Illustrator (which has 64-bit support and all the vector capabilities of Fireworks at the cost of raster abilities and having to retrain most vector operations, they ARE there but the mechanics are different), Inkscape (which lacks a great swath of the plugins I normally use, and also has massive loss of raster capabilities and Live Effects) and just giving up on the project because my tool of choice was insufficient.

    The choice I took was venturing into the very depths of R'lyeh. All Adobe products allow scripting through a Javascript Engine. Javascript is infamous for being a crime against programming. The Adobe Javascript engine for being a crime against Javascript. And Firework's code for being a crime against all sense of decency. While Flash was basically rewritten from the ground up when bought by Adobe as they wanted it as a flagshhip product back then, Fireworks had no such luck and was basically running on Macromedia code to its day of decomissioning... back in June 17, 2013.

    A flattened layer is a simple raster image, raster images are considerably smaller than transformed vector layer, specially when accounting for vector masks. The obvious downside is that all vector information is lost and the object is no longer editable as a vector object... ever. Harnessing the powers of Javascript I wrote about 15 different scripts which had the ignominious task of changing the name of the selected layer/s into a unique identifier (UID), depending on the type of the layer (pure curve, curve with effects, curve with masks and so on) it'd store the information needed for reconstruction in a separate text file and flatten the layer. The complementary script would do the inverse process, get the UID and load the original object and delete the raster version. This meant I could manually unload objects if I wasn't going to be editing them for a while and load objects when I needed to go back and edit. This considerably lengthened the working process as the scripts were written not for speed but to minimize memory usage. An external python scripts (which had the 28GB of RAM Fireworks couldn't access) would convert the javascript text file into a an optimized representation for loading which the loading script could access.

    Another element of this horror show was that Fireworks had roughly three types of out of memory failure. One was due to rendering an operation. this were relatively benign, all your work would be lost and you were likely close to the limit already but it didn't have any further consequence. Another came from objects in memory, this normally meant that the problem would also be present in your last save if it was relatively recent. You'd open the file, move a single object and get an out of memory error. This happened enough times that rotating save files was a need for my mental health, this also meant that unloading would not be enough at times since the javascript engine also required memory to be able to write to file which happened before the rasterization. And the final type of out of memory error was the ones resulting from a save operation, this inevitably made the operation fail, the file it was overwriting get corrupted and general merriment to be had.

    The result of this was another external python script that monitored Fireworks memory usage and would blare at me with Verdi's Dias Irae when it was close to the 3GB RAM mark so I could take appropriate measures. My current art project is a transcription of the Requiem, from memory. I got more aggressive with curve and object unloading after the first week of art work after a week of around 60 hours of pure programming.


    All of this hassle also lead to one of my major regrets, not fully detailing the cliffside and forest layers at the top. Because of the amount of time spent coding and repeated remakings of work from memory failures, quite a bit of this section was not up to the standards I wanted to set for myself and which were given to further sections.

    Spoiler: The Finished Top Section
    Show

    Cloud colouring was a heavily involved process between vector masks, multiple displaced layers of prerendered flat clouds and perspective shifts and edits. The file showing the edges shows how much of it was perspective tricks and the nebula which was ultimately removed in the final version.

    Spoiler: The Top Section's Complete Canvas
    Show

    Planning Round Two

    Replacing the draft section with the finished upper third resulted in major revisions. The first was the canvas expanded upwardly (this would prove a problem later on... seeing a trend here?), I had to accommodate the higher moon position. This meant that the original blueprint was now EXTREMELY lacking. The proportions were massively skewed in favour of the upper scene. Not only that but the colour balance was MASSIVELY affected. The original plans were much whiter, depending heavily on fog light dynamics, which the night sky did not favour. I had originally misjudged the ability to have strong blacks on the objects involved, relying on the night sky proved fruitious, the deeper blues inspired a much stronger palette choice which also worked quite nicely with the introduction of Rennaisance references and motifs.

    Spoiler: The Sketch with Completed Upper Section
    Show

    This required a great deal of reorganizing and basically redoing the remaining sections of the sketch. While the focal ideas remained constant there was heavy rearrangement. Furthermore one of the major themes through the rest of the work was decided. The objective was an image that would good at all level, from 100% zoom to 20%. Both at detail level every section would be designed to have something to look at at normal zoom levels but also to allow and guide the flow of scrolling around and to work in those levels as well.

    I gave considerably more space to each section so it was possible at 100% to 66% zoom to have a single focal scene in sight, but also had to add more elements so each scene could stand on their own. The brunt of the planning work then shifted from the individual sections to the transitions.

    Another Fireworks TM technical fact is that the image canvas is limited to 6000x6000px, this meant that there was a very real limit into how much vertical space I could provide. This limitation had no solution, it was something that thankfully didn't become a major issue. A hypothetical solution exists but thankfully I didn't have to recur to it.

    Spoiler: The First Sketch Revision
    Show

    The Surprisingly Uneventful Mid-Section

    There is little progress snapshots of the mid section on account of how quick I went through it. The work on the waterfalls was manually using the pen tool and painstakingly going through their entire length. Most of the rock effects where vector mask trickery and effects and the character was a single seating. Loading and unloading were second nature (and bound to keybindings by now). And the fog effects where spherical distortion transforms over blurred cloud renders with some retouching of vector layer effects.

    The petals took some attempts to get exactly right. Initially part of my programming goals (which were ENTIRELY DIFFERENT from the programming I had to do by the end of this) was to code a small engine to take pattern objects, a trajectory curve and "blow" the objects along the curve. This math and programming exercise took about two days of discarded work that generated locust swarms of petals before I simply did it by hand.

    At this point I was forced to switch from my mouse to a drawing tablet. Most of the time I didn't stop working because I was done but because my hand pain flared up. The analgesics that kept it down also made me incredibly drowsy meaning I could either continue for the day with an iron rod jammed through my wrist and my fingers refusing to click or I could be three-fourths asleep and advance a vertex an hour. The tablet helped enormously in that regard more than in terms of speed and capabilities (I'm not fully used to it and it takes considerably longer for precise work which most vectoring is).

    Spoiler: The Middle Section Objects and Masks
    Show

    The Second Sketch Revision

    Inserting the middle section into the sketch was a massive relief. While there was some work in making it fit properly on the edges and very minor retouches, this part was considerably smooth.

    Spoiler: The Middle Section Added to the Sketch
    Show

    However, yet again the problem of proportions came up, and this time also the issue of the transition. I was extremely unhappy with how I'd have to deal with the edge of the bottom of the big waterfall's fog. Furthermore the bottom section would have to balance out the top section, details and colour wise. There was a hefty amount of reworking the bottom, this also hit some of the previous transitional elements, the petals blowing into the bottom, the smaller waterfalls and the pieces of continent mass falling apart on the background.

    After being justifiably chided by my much more artistically experienced sibling I did it the old way, drawing the contours as best as I could with a pencil and planning everything in advance correctly.

    Spoiler: The Third Sketch Revision
    Show

    The Very Amenable Drawing of Hell

    I started with the central character and the focal point of the scene. It was relatively quick, except the fact that the light source would prove to be somewhat challenging (it would become CONSIDERABLY challenging when dealing with the landscape).

    Spoiler: The Focal Character
    Show

    After some deliberation and fitting to the theme and who the character was standing in for, I completed the lower part of her body.
    Spoiler: The Mid-torso in Progress
    Show

    Spoiler: The Character's Full Body
    Show

    Besides my constant fighting with the light source there was mostly busy work of drawing cliffs and fighting out how the bottom light source would affect them. I did everything on grey before adding the coloured light highlights.

    Spoiler: The Left Cliffs Done
    Show

    The foreground pillars and the interaction between the waterfalls and the bloodfalls took a bit of work because of the load-unload process but also were relatively straightforward.

    Spoiler: The Foreground Completed
    Show

    And the last overview of the sketch before the final version.

    Spoiler: Uncoloured Lower Section
    Show

    This was my other big regret regarding the final result, I was unable to draw and shade the clouds I needed. The clouds would have had the lower black and red shading and a white top (Like Corona Mountain clouds in Super Mario Sunshine), not only that but they had considerable perspective warps to be able to properly fit the middle and bottom sections. My skills proved insufficient and after an almost entire day of failed attempts at a single cloud with no successful results or even close to them I settled for a simpler solution.

    Colourwise most of the work was done in a big sweeping stroke of the bottom section, the end was in sight and I sprinted to reach the end of it.

    Yet again, the link to the finished version. Unless you have a 4k monitor you'll need the side-scrolling.
    Last edited by AsteriskAmp; 2018-03-31 at 04:34 PM.
    The Iron Avatarist Crypt of Fame - Exorcising photobucket from the historic archives of the forum.
    Go and went by many names Ast, Avgvst, Pink-Haired August, araveugnitsuga and nowadays AsteriskAmp.

  2. - Top - End - #2
    Ogre in the Playground
    Join Date
    Jul 2008
    Location
    Germany
    Gender
    Female

    Default Re: The Flower's Descent - A Big Art Project Post-Mortem

    That's an incredibly involved (and gorgeous) art piece to do with vector graphics! Kudos to you! :O Very nice flow and composition to it, and I zoomed in on some of the details - wonderful.
    *Above post: Additional terms and restrictions may apply.
    My old OotS fanart
    My art on Instagram
    My Patreon

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •