I added Hepmonaland to my now old map of the World of Greyhawk in 2018 which was all done using Photoshop. The last bits of terrain using Bryce 3D was added in 2013, even though I have updated it every year since, the map is now old and in need of a replacement, not more cosmetic updates. It has been my main mission to do this for over a decade, trying to figure out what is possible and how to best go about it. It needs to be fully 3D, highly detailed, and have all the data needed for GIS, game engines and so on. A decade ago World Machine had been recently released; I saw its potential and started my quest to learn to use it. A decade of experimenting, trial and error and a long slog to try and get it to create the vast expanses of realistic landscapes I craved. I was ready to give up many times, both due to the complexity trying to overcome all the issues I encountered, and the render times involved. In the beginning it took me over 24 hours to render a 16K area on the computer I had back then, today, thanks to both better software and hardware it takes a bit over an hour.
During this decade I’ve mapped Midgard for Kobold Press which forced me to get stuff done within a certain time, and I where determined to move from Bryce with the Midgard project. For a year I lived with Southlands learning World Machine, and after a restart 8 months in I scrambled and got a decent continent done at the end of the year. Look closer at it though and there are no small rivers, just the major ones that are not much more than lack of land creating more sea level. For the main parts of Midgard, I refined my technique a bit, but it was still very basic getting stuff done in time. I knew I wanted more for my next set of Greyhawk maps, proper lakes and rivers, and real 3D terrain.
The greatest hurdle to overcome was lakes and rivers, to make them flow naturally and forming drainage basins like real rivers. It has taken me thousands of hours to get it to work, but eventually I figured out workarounds that produced a more than good enough result. The key was to break down the effort into stages and then set up a workflow to manage data sets, masks and templates to make sure quality and efficiency could be maintained throughout a multi-year project to map the World of Greyhawk. The last five years I’ve used Shield Lands in my campaign as a testbed to see what works and how to get it done effectively. The results have been great, but with lots of work, over time I have figured out how to do things in a smarter way which gives me the results with a bit less work. I’ve also, thanks to your support, been able to upgrade my hardware. So now I’m using two powerful desktop computers, both capable of rendering effectively, which means I can work on my main machine, and let my side machine render in the background. I have a third older desktop that I’m using for streaming, email and such, which will be replaced by a new laptop. I’ve ordered a Framework 16 which will come in December. It will be powerful enough to load and preview even the largest of files, so I can show more than screenshots at conventions.
I start with loading my current Flanaess map into World Machine, as a guide. Still prefer WM3 for this stage it is reliable and able to provide stable previews but will use both WM4 and Gaea for special terrain creation that can then be imported back into WM3 and placed in the terrain. World Machine is still the best when it comes to large area terrain creation, but Gaea is closing in there might be a reason to shift to it for this step too in the future.
Key bit here is to set up multiple parallel chains of fractals, various filters and erosion nodes, blend them together, make tweaks when needed and add extra erosion here and there to increate variation and realism. I’m including lakes and rivers in the previews at this stage to make sure that the terrain is leveled in the right way to match existing maps, but no rivers will be included in the resulting Stage 1 heightmap. For this stage I realized (after a lot of trying) that tiled renders didn’t work so I was limited to 16K (16,384 x 16,384 pixels) in size and making it cover a bit over 100 miles gives me roughly 100 feet per pixel. Not perfect but good enough detail for a base heightmap that can then be further refined later. When I have rendered an area, I’m importing it back in World Machine when doing the neighboring areas to try and make them fit together. This means I might have to go back and forth and re-render, import back, adjust and re-render again until they fit together. It is a process with some pitfalls but I’m getting better at it and have now covered a sizable part of the Flanaess from Furyondy to the Griff Mountains
When the base heightmap is done it is time to use the power of World Machine 4 and import several base height maps covering an area over 500 x 500 miles. This makes it possible to do a tiled render that is 64K in total which is a great base for local area and settlement maps with a much higher level of detail. Keeping records of exact location of these areas, file and directory naming schemes are vital, often half a dozen renders or more are needed to get a good enough result, so version marked files and routines for where to put things are needed, as well as purging the leftovers to fee-up hard drive space. Each 16K area render is around 6GB, and now I have 26 of these areas either done or well under way, with all the renders done a bit over 300 GB for Stage 1.
Even with my best efforts in step 1, the different Base Heightmap are not going to fit together perfectly, so before full resolution tiled render two steps of error correction and refinements are needed. Blurring, using fractals to add new terrain features to hide the ugly bits, erode again, and in some cases just lower the terrain to adjust coastlines are the tricks needed at this stage. If nothing else works, Photoshop editing of the heightmap is the last resort. After a day or two of work a dozen or so 16K areas have been successfully blended into a single heightmap, more than 500 x 500 miles ready for the next crucial step, water.
This is a very interesting part of map making when the landscape starts to come alive, with lakes, rivers and coastlines. It is also very frustrating, rivers have a tendency of flowing in direction you don’t want, and lakes rarely form where the source maps have them, so tweaking the terrain might be necessary. In extreme cases it might be needed to go back to Step 1 and redesign an area again from scratch, but most of the time all that is needed are slight tweaks here and there to get things right. When I’m happy with the preview I set up exports for all the masks, sea mask, lakes rivers, a separate height map for water depth and a simple colored weight map that can be used later to help with texturing.
At this stage it is time to extract and edit all the data, height masks created so far and then generate all the additional masks needed for texturing and detailed maps that I want to create using this base data, because it is only a collection of height data and masks at this stage. I set up additional erosion devices of various types like coastal, water and wind based, not to change the terrain but to use the data from them to generate flow, wear, soil and deposition masks. This will guide texturing by providing data for placement of vegetation, fallen boulders, bare rock and more.
Now it’s time to take a closer look at the edges and overlapping areas of each of the full resolution tiles, this time it is mainly lakes rivers that are of interest. The tiling makes lakes that cover more than a single tile often run into problems with shifting surfaces that need some TLC to establish the correct even surface height and extents. Importing all the data into a new WM4 file makes it possible to see the terrain in 3D to make sure it looks right. Rivers often have similar issues across tile borders which also need to be blended, either in WM or by painting on the masks using Photoshop. Very similar to the correction of Step 2 but less drastic since at this stage I’m working with already rendered heightmaps, not fiddly fractals.
A special case are rivers that flow down steep highlands and escarpments, which World Machine always makes flow down along the slope of the cutting into it evenly the whole way. Sometimes this is natural and the right thing, but in places water would erode it into the slop in a more step-like fashion forming a waterfall. This needs to be done manually using WM and at times Photoshop, which makes rivers look way more natural.
When all the errors have been fixed and the data extracted in the forms of height- and depth maps and masks, it is time to bring in the colors. This is my favorite part when tings truly come to “life”. It is difficult and often takes several attempts, but trial and error that are setbacks at first often leads to new discoveries that can be used later for different areas. The colors of a world are closely tied to its climate, geology and the creatures living there, so when you have a set up for one tile, it neighbors can use the same setup, just feed the different in-data and then do small tweaks to compensate for latitude, closeness to large bodies of water, rain shadows and other local conditions.
To be sure I render a lot of texture variants each, for rock, water, flatlands, highlands that each cover everything. Then I use the mask data in Photoshop together with the powerful blend modes, opacity and further tweaks of color, light, contrast and more. To work with textures in Photoshop makes it possible to work in real time at full resolution, which is powerful. When things don’t work out a bit of manual editing here and there is always needed, and even preferred, it makes things look more varied and artistic, which in my opinion improves beyond what fractals and procedural can achieve on their own.
This base map and its data will serve as my new base to work from. Whether it is coastlines and major rivers for a simple spread map, or all of it for a city map, it will be very useful to keep consistency and speed up the result. I’m striving to use standardized data formats as much as possible so what I’ve created now can be useful for a very long time and in many different ways, in Photoshop GIS, on the web, a VTT, to be printed and in game engines.
My goal is to keep Greyhawk the best mapped fantasy setting in the world, we are not there yet. Both Midgard, Middle Earth and Forgotten Realms are competing for this title and in some ways ahead of Greyhawk, but there is hope that we can change that in 2026 and 2027. With your help it can happen, and this blog post goes over the process a bit. I’m working hard on a part of the Flanaess for the Melf’s Guide which will be a first use of this effort, whish is a dream commission to work, but still only a minor part of this project that I plan to work on for the next five year. Then I hope that the base map of Oerth is available for everyone who wants to download and use it.
There will be deep dives into each of the stages here in future posts, and if you want to see it in action. Well action is to push it, it’s more like watching paint dry, but if that doesn’t deter you please stop by my Twitch Channel https://www.twitch.tv/anna_b_meyer on Friday evening @6pm PST.
Thank you again for all your support!!
