JAY TARRANT
VISUAL DEVELOPMENT | CONCEPT ARTIST
VFX PATHWAY
UNIT 8 -
VFX - TV SHOW ROOM VR CREATION
At the briefing for this project, I was immediately drawn to creating SPECIFCALLY the computer room from Lost, and for a good portion of the project that was my complete focus. I thought this could be quite an iconic environment to work with, due to the high amount of content available within it, and also being that 'The Swan' is the connecting point to the survivors and the island, with the string of numbers '4,8,15,16,23,42' being in its greatest use here.
​
Just before the Easter holidays, I realised that being confined to just the computer room, didn't give much of a 'VR Experience'. I decided to expand my idea outside of the computer room, to the rest of the swan.
​
During Easter, I set out about modelling as much of the Swan base as I could, to give myself some wider opportunities to make this an interactive experience, and make it more interesting. Doing the entire base gave the whole 'experience' more of an eerie experience, by having more to explore than just a spherical room. Hallways that lead around in a maze-like manner, darkness, and odd noises. I'm not sure how much of this I'll be able to pull off, due to not knowing much about the confinements of VR, and the large amount of space I am working with.
​
Currently in time, I am only planning on making the computer room explorable, or viewable, whatever, the other two rooms that connect to the computer room will be partially seen, unless I zoom through my initial idea, and push myself to make one or both of the other rooms explorable.
​
The first two floorplans above are fan-made floorplans I found online, these are much more detailed than the production ones, with the one on the right hand side also including the tiling patterns seen on the floor of The Swan.
(https://lostpedia.fandom.com/wiki/The_Swan)
​
Below are two of the production floorplans from the show, one pre-explosion in the swan (this was sold afterwards on a 'Lost' collectibles website, and the other I found online from the Production team information.
​
​
Finding these floorplans was extremely helpful, we are frequently looking at the room from behind the computer, seeing the machinery and equipment, and the computer obviously.
We don't regularly see it from the odder angles, so working out the placement of objects in the room was becoming quite difficult, however this now gives me a roughly exact look of the placing and whereabouts of all of the items found in the Computer room.
(and also the floor tiles, which honestly I wouldn't have even known were different colours in certain places if I hadn't found these floorplans.)
BREAKDOWN OF THE COMPUTER ROOM
27-02 UNREAL ENGINE SHENANIGANS
In class we were lead through a brief introduction to the BLUEPRINT element of UE, the 'node/coding' system UE uses.
​
During class, I decided to continue playing about with what BLUEPRINT could do, extending slightly from what Molly showed us, which was originally showing us how to use the player to trigger events. I.E player walks into a certain space, and a light will turn on.
​
The above video is a brief walkthrough of what I managed to create. A system in which the player has to shoot cubes to turn lights off, after the player activates them when walking into certain spaces.
MODELLING THE DOME
For a worrying period of time, I struggled to model the dome. For the first few weeks of the brief, the dome was my make or break factor, as it is the most prominent part of the set, If I couldn't successfully built the dome, it would ruin the entire experience.
​
Starting to model the dome, I tried many different techniques, attempting the model the triangles flat, and then grabbing the middle to create a dome. This resulted in very stretched out triangles, and it didn't look good.
I then went on the create a sphere, and manually attempting to edit the vertices myself, this took ages, and also didn't work (it probably would have, but it took so much time and effort I scrapped the idea)
Then, after doing some research, I realised blender actually has a add-on for Geodesic Domes! I tried this, and it still looked weird? I also couldn't figure out how to scale down the width, also it wasn't technically a dome, as the dome was made up of 3D shapes, the inside of the dome was actually quite square. you can also see this from certain points of the dome, where they meet in a square shape.
Finally! I figured it out, I found a tutorial on YouTube of someone making a dome, and yes they did use the blender add-on, but they also created the dome with an ISO-Sphere, not the UV-Sphere normally used. This worked perfectly, with some extruding and bevelling, I successfully had a dome.
MODELLING THE COMPUTER
Modelling the computer was actually quite simple. I originally tried to model the airflow holes on the computer and drive, before I remembered about the Boolean Modifier blender has, using this made it 10x easier. All I had to do was model the shape of the cut outs I wanted (much easier than the reverse) and use the Boolean Modifier too cut these shapes out of my existing model.
​
At this point I haven't finished modelling the back of the computer/monitor, as I'm not sure if this will actually be visible in the build of not, however I have given myself a good base to add these additional details, along with the wires, later if I do decide that the player can roam.
The keyboard was probably the most detail oriented modelling process, however it was still quite simple. On the Computer History Museum, I found a full 3060 turnaround of a Apple II Computer.
Using the picture above, I used this as a template to layout all of the keys I needed, and then extruded and scaled these to successfully make the keyboard.
​
Currently this is all modelled as one object. So I need to be mindful that if I do wish to have a 'interactive' keyboard element in my experience, this needs to be separated and animated correctly.
MODELLING THE DESK
Creating most other objects now, was relatively easy, as there was lots of duplicates of objects in the room. For example, the stack of 9-track disks on the desk, All I had to do was create one track, and then duplicate this and rotate it a little bit. Easy, 6 'unique' 9-track disks created in no time.
​
I had a similar experience modelling the floppy drive, using all the techniques I have already mentioned above,
​
​
MODELLING THE OTHER ASSETS
Modelling the tape tracks, and the SAGE, was really quite difficult. There are lots of little details in the model, that even after simplifying the design, I was struggling to model. Luckily, I had done research into all of my assets, so I had measurements for all of the machinery. This made it significantly easier when trying to understand the size of each object.
I ended up changing the model of the SAGE machine, quite a few times, by the time I finished my environment. Texturing ended up becoming a very big problem, so I removed more elements of my model, and opted to draw some assets. Having the knowledge of how this project would pan out. I wouldn't have made this decision.
​
For a very large amount of this project, I was very worried about how my environment looked in Blender. Although this did make my life easier when transporting my scene to Unreal Engine. If I had known that I would be retexturing my entire scene, as well as re-doing all of my lighting. I would have kept a lot of my initial models, instead of trying to simplify them for Blender's texturing system, that is completely different to Unreal Engines.
While looking at some online resources, I came across a method of modelling and texturing to make my life much easier.
SHIFT+D will duplicate anything selected directly. SHIFT+L will allow me to link UV maps and materials to separate objects.
​
However, making my life so much easier..... ALT+D, this links UV maps, materials, objects, and treats the duplicated object as a child of, kind of, but if I changed the original asset, every object I duplicated will also change accordingly.
06-03 UNREAL ENGINE TESTING
Trying to load what I had modelled so far into UE for the first time, was not successful.
​
My models were wayyy too big, which I knew already, but regardless I wanted to test them outside of Blender.
However, as well as this, when I tried to rebuild the lighting, I got a big string of error codes about my model. So the lighting wouldn't render correctly, which meant I had a quite odd looking blank model.
​
Regardless, It was very cool to be able to see my model outside of Blender.
After researching online, I realised later on in the project that these error messages were quite a tedious, but simple fix. A lot of my UV's hadn't unwrapped properly or my textures were much bigger than the texture map, but because of the way blender tiles its textures, I hadn't actually noticed, I had to go through my scene and unwrap everything again. (Only a week later to get told I could unwrap everything at once...)
BEGINNING TO TEXTURE
I spent a really long time messing around with lighting in Blender. I didn't need to, as none of it would transfer to Unreal Engine. I knew this to some extent, but still continued to worry about it regardless. I wanted to have a base idea of the type of Ambience I wanted in my final scene. It did come in handy while using Unreal Engine, as it gave me a direct reference of the outcome I was looking for. But It took aa lot of time off of my schedule on something that ultimately didn't affect my scene. I also hadn't researched in full what textures and materials would transfer across to UE5. One of these was emission materials. Luckily for me I had only created a few materials using emission. However it was annoying to have to redo these.
PAINTING DESK TEXTURES
Originally, I started off trying to export my UV'S into photoshop, painting 0ver then with textures I deemed appropriate, and then importing these back into Blender.
This caused me quite a few issues, with detail and also the accuracy of the UV maps (even after unwrapping).
​
In the second image, you can see an example of the texturing I wasn't happy with. The computer does have a very cool illustrative style to the model, however this wasn't the style I was looking for, and I actually felt that it didn't look illustrative, it actually looked like lazy texturing.
​
What I decided to do instead, I believe actually worked quite subtly, but very well. I would take the original texture, whether this was a flat colour, or an imported texture. Using the texture paint feature I'd paint over the model, using the fully rendered view as reference, to accentuate shadows and details caused by the lighting, into my models texture.
I ended up not actually 'texturing' my models much, all I ended up doing was accentuating the details that can be seen in the rendered version.
​
Comparing the first and 3rd image, you'll notice that there isn't any more texture on the model, all I have done is accentuate and widen the shadow and highlights already seen in the first image, but adding this detail into the models texture, has made a massive difference to the texture of the computer.
A little bit later down the line, I realised I completely forgot about the bowls on the desk. Although it isn't much, I thought it was quite important do add as many small details as I could, especially as they weren't difficult to make. From the first looks at my environment, with the basic props in it, its very empty. It doesn't look particularly weathered or 'lived in' for an unknown period. My plan was, once I had modelled the essentials, to go back around my environment adding filler objects and 'tat', rubbish bits. Just to make the whole environment seem a bit more used.
OSCILLISCOPE TEXTURES
I decided to paint the text and details on the front of the oscilloscope, it seemed easier than trying to project the textures on, especially as I was trying to minimise the amount of nodes being used on my models.
​
The drawings were a bit rough, but apart from looking stylistic, compared to the rest of the scene. I think they came out really good, and I actually like the drawn style up close.
​
The oscilloscope would only be seen from across the room, so it mattered more to me that the lines were smooth and focused, rather than looking printed, but blurry. I added a point light in front of both scopes screen, in order to imitate the light it would have been emitting
I need to change the design in the radio waves bit. As although it is accurate to the design of the oscilloscope I was referencing, it isn't actually show accurate, in show this is actually green radio waves, I am unsure right now as well if I am going to animate the radio waves to be moving.
SAGE AND TAPE RECORDER TEXTURING
Texturing both the SAGE and the tape recorders was an absolute nightmare. I started out originally trying to separate each portion of the model, texturing individually. This turned out to be awful, because weathering and 'destroying' the textures was taking ages, as I now had 12 or 13 different objects to texture instead of 1.
​
In the end, I ended up painting over the textures, getting all the detail that way. It does look very stylistic, not particularly in a good way. From afar it doesn't look bad, however compared to the rest of the texturing in the room it does look half arsed.
​
The lighting systems took a lot of effort. I realised that emission modifiers more than likely wouldn't transfer successfully to UE5, so I added individual point lights to the front of the models, which worked a lot better, and actually made the light system look better. I also knew this would transfer to UE5. Doing this also meant it would be easier to animate the lights flickering, as each light was on an individual scale, rather than being connected to the model.
​
I also remembered that later on in the season, the machinery in the room is covered in tarps, I thought this would be a good feature to include in my environment, as again, it made the room look more used, rather than a brand new set.
REST OF THE ENVIRONMENT
Once I finished the computer room, I struggled to think of a way to block off the entrances, without seeming like the player was 'trapped' (even though they were). I decided originally to just build the first section into the rooms, to make it seem like there was more beyond the door.
While doing so I decided that maybe I should open up these rooms to my VR Experience, as my experience was currently quite lacklustre.
I didn't really put together how difficult doing these other rooms would be. From first glance, in the show, the rooms do look mostly square. After actually beginning to model them, I realised this absolutely was not the case.
The pantry was quite a model heavy room to build. Luckily, everything was essentially built up of the same 4 or 5 objects. So using the Boolean Modifier I managed to quite easily build the shelves, and then, creating some glass jars and cans, then changing the products in side and the labels on the side, I had a million different products stocked in the pantry.
One big positive of doing Lost, which I realised very quickly, because it is a show with such a MASSIVE fanbase, almost anything I wanted to make or reference, had already been made as merchandise. All of the product labels had been mimicked by fans. All I needed to do was find some of these images, using them as reference to make my own labels, and then use them as textures.
The Dining Nook was another relatively easy area to create, as I reused most of the assets from elsewhere in the scene, apart from the walls, every wall in this built caused me so many problems, with them not being thick enough and letting light through, to overlapping vertices causing glitches in the mesh.
I decided to reuse a model I had made for Unit 4, the blinds from a bedroom, purely because, why would I model it again if I have the asset ready to go from a previous project?
I absolutely loved the way the light filled the room in these renders, When exporting to unreal Engine I tried and tried to mimic this look exactly, because it was so pleasing to me. But unfortunately I couldn't completely emulate it. I am still happy with the lighting in my UE5 file, however the lighting in this specific section is disappointing I couldn't recreate it.
LAVA ROCK ENTRANCE
This entrance was made from just one rock. That I duplicated what felt like a million times. I made a mistake at first, as I had joined all the rocks together, meaning that my UV maps, without overlapping, were absolutely useless, as there was so many rocks projected onto one UV map there wasn't enough space for them all.
​
This was the first walkthrough I had done of my environment. Although severely under finished, I wanted to get an idea of the experience so far, as I hadn't exactly planned for my project to blow up so much. I really liked the flashlight element I had decided to add, after talking to others for advice and recommendations. The flashlight meant my experience could be much darker than what it originally was, and it matches closer to my original scene inspiration in Lost.
THE MURAL
By the time I had finished texturing, this mural was my worst enemy. I researched so many different solutions and asked so many different questions. I wanted to use the ORIGINAL mural from the show. I knew I could not do justice to the original mural if I were to repaint it.
The issue I was facing, was getting the mural to overlap the original wall texture. I tried using a transparency method, which didn't work, It just made the mural barely visible. I created an Alpha image, to see if I could get that to work, and get Blender to ignore some of the Mural. No Luck. Finally I found a really old Reddit post, talking about Blender 2.9 or older. The Reddit post mentioned a method where instead of combining two image files, I combined two separate Principled BSDF materials. IT WORKED! mostly. It was the best attempt I had had so far. I turned down all of the Specular and other "shiny" aspects of the principled BSDF, and I managed to get a really close match to the mural.
​
I decided to create a second render walkthrough, to see my progress so far, with new assets and more textures. I was really happy with the results I had so far, and the more objects I added into the scene, the better it began to look.
FINAL BLENDER DESIGN
I had made a lot of progress very quickly in my scene, I wasn't able to document the timeline of every asset and room creation in my scene, a lot of new objects appeared, including a good few I wasn't planning on creating. After showing my original two walkthroughs to friends and family. A lot of people bought up the exact same criticism every single time. its too clean. The Hatch would never be that clean. So, I went through and retextured the floor, walls, ceiling and a couple of assets creating a dirtier environment, which after reviewing again with friends and family, was much better. Unfortunately, in retexturing the walls, I lost all progress I had made in making the mural slot in nicely with the wall. I left this, as I was extremely close to exporting into UE5, and at this point I had realised that the mural was going to be the least of my problems, and I would research this once my model was successfully in UE5.
LAVA ROCK ENTRANCE
There are certain elements of my Blender renders that came out better than the Unreal Renders. Most of these revolve around textures. In the Entranceway, when importing the Unreal, a lot of the UV maps became impossibly broken, as there was so many vertices and angles in this one section. A good portion of semi-visible sections became stretched out and skewed, and because of how vertices dense this section was (my own fault for joining a bunch of rocks together to create 3 walls) I couldn't find a way to re-UV unwrap it to make the texture look as good as it did in Blender.
FIRST HALLWAY
SECOND HALLWAY
SIDEROOMS; LAUNDRY ROOM, PANTRY
The pantry was another really weird transfer to Unreal. For some reason, the glass didn't work as expected on import. Even changing the texture to a different glass texture, it went slightly opaque, and wouldn't let the objects inside be seen properly from the outside.
​
Even in Blender, The pantry took such a long time to model. Although it is only the same few objects duplicated, placing them and duplicating them took a long time, and it still isn't as full as I would have liked. Unfortunately the pantry was just too heavy to keep adding assets into, it almost completely crashed Blender as its own isolated room.
LIVING ROOM AND DINING NOOK
COMPUTER ROOM
The computer room was the only room in The Swan that I had drawn most of the textures for. Once it was finished it looked really good, making everything light up the way it was meant to was a nightmare, especially after I forgot that Unreal doesn't like emission textures, and I had to go through and add point lights to all the machinery (it worked out to over 100 point lights in this room alone).
​
Although it was a nightmare, I'm really happy with the painted textures. From super close up they don't look super detailed, but from afar they make the room look really complex.
IMPORTING TO UE5
By the time I successfully managed to import my scene into Unreal Engine, I was on my 7th or 8th attempt. Throughout the class, there had been a lot of issues importing our scenes. Speaking to Poppy, we were researching a different way to import our scenes into Blender. Using the USD Stage method we had been shown, led to an extreme amount of glitches in our textures and models, if it even managed to import. After doing research and looking into my options, I tried importing my scene as a .FBX file. It worked better. On successfully importing my scene, I lost all my textures. It meant I had spent a lot of time texturing in Blender, for what felt like no reason. But the scene had imported correctly, I would rather have all the components of my scene than the other way round. It wasn't too awful retexturing my scene. I had to reimport some sections, like the oscilloscope and computer, as on import UE5 had changed my UV maps, and these were painted textures, so I needed the same UV maps as Blender.
While re-adding textures. I thought I was missing a lot of my scene (an almost devastating realisation that I hadn't fixed my problem by using a .FBX file instead of .USD). I figured out that a lot of my normals had become flipped on transfer to Unreal. Using the Modelling tools plug-in for Blender. I had to go through a lot of assets and flip the normals in certain areas of my models. The Modelling Tools add on was probably the most helpful plug-in I used in Unreal. It helped me rescale and re-wrap by scene, as well as separating my model by texture and de-stretching/ removing elements that had messed up on import.
INTERACTIVITY ISSUES
One of many issues I faced with the interactivity in my experience was the flashlight I had made. I wanted the player to be able to pick the flashlight up and then carry it with them.
I couldn't get this to work.
​
I tried simplifying my idea, by making the flashlight disappear once the player had made contact with it, in exchange for a beam of light, mimicking a flashlight without the constant view of the flashlight.
​
Couldn't figure this out either.
​
In the end I settled for an auto-equipped light, that had no interactivity whatsoever, apart from following head movement to light up the space in front of the player, sort of like a flashlight.
UNREAL LIGHTING INSPIRATION
I had already been playing around with my possibilities for lighting, in both the concept art, and while texturing in Blender. Ultimately it didn't matter much what I lit until it was bought into Unreal. I created a mood board to inspire my lighting design. I knew I wanted this very isolated, spooky feel within the scene, however there were many ways in which I could have explored that. I really liked this cold, foggy feel. It gave me this assumption that not many people had explored this area in a while, as if the player shouldn't be there. I really liked the feeling these images gave off, as although there was nothing inherently scary about them, they all had this very spooky vibe to them, which is what I wanted to channel somewhat for my lighting design.
.png)
FINAL UNREAL RENDERS
My final outcome definitely looks a lot better than how I expected it to. With all of the technical difficulties we had, alongside learning a completely new program.
I didn't manage to do everything I wanted to, however I'm also surprised I got as far as I did.
​
The ambient fog and dust was actually a last minute shot in the dark decision. Which was an odd decision really, as all of my inspiration pictures included all of these elements, I guess I just hadn't thought about it. These physics elements really elevated the ambience in my environment. I was worried before, due to it being overall really dark, but also too bright... and too flat.... the fog helped break up the environment and make it appear denser than before.
I had a bit of issue working with the flashlight and fog by the end. The flashlight looked like it was coming through the fog too harshly. It was probably accurate for the density of fog I was working with, however I did not want to turn the fog down as I liked the way it affected my environment currently.
I resorted to turning down the brightness on the flashlight, which made my environment darker than I wanted. I solved this by placing a secondary spot light behind the player, to almost make the player an additional flashlight, without it being obvious that the scene was being lit up outside of the flashlight.
​
I was disappointed that I couldn't recreate the light shadows I had managed to create in Blender, however it makes sense as they are two separate programs.
​
In the future I plan to go back and play around with the lighting some more. During my research for this project I found an article talking about different reflection point of lights, to make them appear stronger and project out in more unique ways, rather than just a general bulb glow I have in my scene right now.
This could really elevate my scene, as a lot of the ambience from my environment does come from these dim lights, so having some interesting reflections could increase interest.
​
​
​
Overall, there was a lot of issues throughout this project, and I'm sure my outcome would have been very different without these issues, however, the transfer from Blender to Unreal really gave me an opportunity to experience a new program, as well as experiment with different techniques and styles that I wouldn't have been able to achieve in Blender. I've learnt a hell of a lot on this project, most of it being what NOT to do in future when texturing and modelling between multiple programs.
FINAL UNREAL WALKTHROUGH
I did my best attempt at some sound design for the final walk through. When testing with multiple different audiences, I had a few issues with the sound, as my home computer has a wildly different sound setup to most standard phones/ computers. Using Premiere Pro, I did some audio editing of the song 'Make your Own Kind of Music' featured in the first scene of The Swan in Lost, I thought it was almost mandatory to have it playing. Using premiere pro I edited the sound to be a bit more rough, as if it was coming from an old speaker (canonically a vinyl player but I'm not nit-picky), this was then edited to come from one specific section of the environment, the living room. On top of this, I also added some machine whirring sounds into the computer room, as well as some footsteps walking while moving.