Between the years 1838-1847, John Tallis published 88 booklets with streetscape views of London:
Before visiting a new neighbourhood, intrepid Londoners would use these as guides. Each booklet street facades, a mini-map of the area and a realistic illustration of a landmark (all drawn by Charles Bigot).
The booklets were very cheap because Tallis stuffed them with advertisements.
Each one contained about 5 pages of business ads. It’s likely that Tallis also charged merchants to have their business’ name engraved above their store on the street level.
Tallis’ booklets were like an early “Google StreetViews” and I posted a link to a scan of all 88 to a website for curious technologists. What happened next made me go down a rabbit hole 🐰🕳️
There was a lively discussion and user fritzo said the following:
Feature request, can someone integrate this into the DOOM engine so one can navigate 19th century london?
Yeah, I wasn’t going to do that.
I’ve never made a 3D environment.
Never even made a DOOM map when that game was popular (hi, it’s me, your grandpa).
But the idea stuck in my head.
And stayed.
And stayed.
And, I thought: “how hard could it be to make rectangular slabs and skin them with the images from the maps and put them in an online 3D engine and maybe pop a catsmeat man in the scene?”….
Well, Dear Reader, here is your chance to enter a proof-of-concept 3D version of Tallis’ Street Views:
Delightful on desktop. Monstrous on mobile.
Making the 3D streetview
My idea was to texture-map Tallis’ images over simple rectangles to create a “street”. I started looking for a 3D engine that was free, could run in the browser and would have a simple First Person mode to permit walking around.
I settled on the the CopperCube 3D engine. A simple and easy to learn game engine that could publish to the Web.
Next, I had to find the highest quality scan of one of Tallis’ views. Then cut it up into rectangular sections for each side of the street.
The best image scan was booklet No. 45 from a sale listing at Barry Lawrence Ruderman Antique Maps Inc. The image is set up as an OpenSeaDragon zoomable image and comes in at a huge 53 MB.
One hitch with zoomable images is that you can’t simply “save as” to your computer. That’s because they’re made up of hundreds of separate images that your computer stitches together on the fly.
Thankfully, a tool called Dezoomify let me combine the zoomable tiles and download them as a single PNG file.
After editing and cropping the streetview, I went ahead and created a proof-of-concept scene in CopperCube. It’s imperfect – like “the walls don’t align” imperfect – but it was fun to make. In addition to the street facades you’ll see the map of the area, a detailed illustration of a particular business, and several orbs that open webpages with information about those specific locations (make sure to “allow popups” when they come up). If you haven’t checked out the 3D experience yet, go ahead and click below:
If you were to expand this kind of 3D environment, you could add historical characters and their stories to the scene. You could also lay out streets according to their actual geometry on a map, using the real width of the pavement and real height of buildings.
Personal observations
Working on this project, one of my takeaways is just how sterile Tallis’ street illustrations look. They’re very focused on the commercial life of the buildings. I get it – these booklets were financed by advertisers, and they wanted to present a clean view of life. The real life on the streets of London would’ve been brutal, smelly and dirty. We’re talking the people wearing every single garment they own at once kind of dirty.
During the online discussion of Tallis’ Street Views, one talented technologist ran the etchings through an AI colourizer. Here’s an example of a colour output:
This is an interesting use of AI and really livens up the landscapes. However, if you wanted to make an accurate representation of London buildings, you’d have to show the colours of the actual material/cladding. Note how, some of the other colorized images in the album show excessive numbers of trees on the street and others show the kind of stucco-clad buildings that are typical of Florence, not London.
In my experience, off-the-shelf AI colourizers mangle the Tallis etchings and alter the image so it only loosely resembled the original. Here’s a before-and-after example of this heavy alteration:
I heard a lot about AI tools that take 2D images and turn them into 3D models. I tested out several of them.
A reliable and simple tool for turning 2D pictures into 3D is the “ZoeDepth” algorithm that you can try on Hugggingface: https://huggingface.co/spaces/shariqfarooq/ZoeDepth . It works best on pictures and photorealistic etches. It did not do well on the Street Views engravings.
There’s an cool related animation you can generate with Immersity AI, but it doesn’t let you export a useable 3D model:
Most of the AI tools were flops – there is a lot of misinformation about AI’s capabilities.
The most capable tool of the bunch was Meshy.ai (refer-a-friend link!). The tool is optimized for generating characters and objects, rather than “environments” – so it failed when I fed it a whole street facade with dozens of buildings:
When I fed it an image of a standalone building, it did much better:
Note how the AI added in an extra set of windows and columns on both sides of the entrance (3 in each row, rather than 2). This model is not quite true to the original.
I tried its neat “generate a texture from text prompt” feature. The result was underwhelming. Here is my prompt:
The walls of the bottom floor are made of limestone blocks. the walls of the floors above are finished in stuccco, while the columns are made of light marble. This building is from London in 1840. Photorealistic style, high quality.
And here is the output:
You can see that the model latched onto “blocks of limestone” and clad the whole building in them. Meshy ignored directions to add stucco & marble on the upper floors.
I decided to keep the facades as flat images in the 3D world, and o add the above “Town Hall” model as a toy object in the virtual environment. Unfortunately, it had an extreme level of detail (265,000 mesh faces) and came in at 27MB. Meshy’s paid tier lets you adjust the complexity of the model, and there are also standalone tools that let you simplify a 3D mesh. But I’ll leave that as an exercise to others. I just opted to embed an “orb” link to the finished model on Meshy’s site.
[Edit: I also used Poly.cam to generate a 3D model of the same facade – it generated an accurate model with no hallucinations]
[Edit from June 14, 2025] I tried generating a 3D object from the same facade with AdamCAD. Both the “Max Quality” and “Speed Demon” modes missed certain elements. I actually found the lower-quality model more accurate, and it looked like it had fewer vertices/embellishments (ex. it kept the back of the building as a plain wall) – which was better for my purposes. The .OBJ exports seemed to have something wrong with them – the simpler model was an 86MB file, while the more complex one was 4MB. However, the 4MB file looked very low-resolution in Copperlight.
Finally, attempted to create “street characters” with Meshy to place into the 3D environment but they weren’t quite right. Finessing the models would take skill and time that I don’t have. Here are some of the funniest/worst options from that attempt:
If you’re thirsty for more
I already linked to a huge image of all 88 Tallis booklets. But the David Rumsey site also has zoomable images of individual streetviews and ad pages.
If you want to see how a particular part of London looked in the past, Matthew Sangster created a map that shows where each of Tallis’ booklets is located. Once you find the number of the booklet, you cna find it on the David Rumsey site or on the Internet Archive.
There is a blog where someone does in-depth research into specific buildings that are featured in Tallis’ booklets. What went on in the building, as well as family history and dramas. I highly recommend it. But beware: this is a deeeeeep rabbit hole.
For another resource that covers specific Street Views houses, do a search at the Survey of London website.
If you like the intersection of technology and Victorian London, check out how this Victorian Room was rendered.
If you’re still reading, then you’re a prime candidate for the hard stuff – London Labour and the London Poor. Here is Volume 1 of 4 that you can get online for free (or, if you’re like me, just buy the whole series; abandon it unread; then give it away to a friend).
