Nick Gellatly
Finding a Visual Style
This week had one job: decide what this thing actually looks like. I went in thinking that was a research problem. It turned out to be a taste problem, and the two are not the same.
Starting with research
I began by asking both ChatGPT and Gemini for visual directions for a 2026 re-imagining of My Beautiful Laundrette, with three clichés ruled out from the start: VHS-glitch eighties nostalgia, generic neon synthwave, and the literal Union Jack.
They came back with six directions each. Twelve ideas, and far too many to work with.
So I reduced them to a single page of A4. It turned out the two sets largely overlapped, which was reassuring. Twelve became five real choices: glass and reflection, degraded photocopied print, bureaucratic paperwork, cloth and dye, and a hallucinatory two-worlds-at-once option that both models rated hardest and Gemini told me outright to avoid.
The one-page reduction was useful and it showed twelve directions is impossible – it was easier to look at fewer options. The two models disagreed in an interesting way. ChatGPT argued from art history and was strong on why an image means something. Gemini argued from chemistry and kit and was strong on how to build it. Neither was wrong. They disagreed because the brief was genuinely open, which is the point of Week 4.
The style I built and then binned
I committed to a hybrid: glass and reflection for the structure, degraded print for the contents, with a cyanotype blue bleeding into magenta. The governing rule was the one I had written in Week 3, begin in social realism, let the inner world leak out.
It was a sound argument. The research supported it. I generated a benchmark image to that specification.
And I disliked looking at it.
The first attempt came back as a glossy black sci-fi portal. The second overcorrected into a clean modern appliance. The third was actually good: flat fluorescent light, grimy enamel, a blotted newspaper, rust and limescale, one thin bleed of magenta at the edge. Technically it did everything I had asked for.
It was also bleak, grey and not mine.
I had reasoned my way into somebody else’s taste. The grit came from two AI briefs and a rule I had written down because it sounded rigorous. Nowhere in that chain did anyone ask whether I actually enjoy looking at grey.
What I actually wanted
A few days earlier I had made some paintings in Gemini while thinking about something else entirely. Flat gouache and acrylic, every surface shattered into angular facets, saturated colour, bold indigo outlines, flattened perspective, checkerboard floors.
Looking at them next to the grey benchmark, the answer was obvious. And annoyingly, the faceted style sits much closer to the Political Pop and modern Indian art I had already identified back in Week 1 than anything the AI research produced. The material was in my own notes from day one. I just talked myself out of it.
The transformation survives the change. It now runs from fractured to whole rather than from grey to neon. Early frames are muted and broken into shards. Later frames resolve into saturated, harmonious colour. Same journey, better paint.
The benchmark and the style board
I rebuilt the benchmark in the new style, focusing on storyboard frame 4, Britain going into the wash.
It took three attempts again, for a more interesting reason this time. The first was legible but static, a tidy pop-art poster. The second had proper brushwork and a real anticlockwise vortex, but the contents of the drum had dissolved into soup. Beautiful, and telling no story at all.
The third put the first version’s contents inside the second version’s spin. Flag shard, newspaper, terraced houses tipped on their side, whisky bottle, banknotes and a block-print panel, all whole and readable, all turning. The left quarter kept flat and empty, because that is where the credits go.
From that one image I built a style board with everything on it: palette, typography, environments, characters, recurring motifs and the rules.
Every colour on the board was sampled from the benchmark rather than invented. Ink navy #121D27, laundry cream #F0CC96, Powders orange #DE802D, vermilion #C0421A, drum turquoise #3F7D75, deep teal #2A5350, ochre sand #D5AF75, brick brown #663E26 and pale sage #89A99B, with turmeric and coral held back for the later frames. Every motif tile will be cropped from the benchmark too, which answers my own Week 3 note that a style board should not become a scrapbook of things I like.
Typography
Poppins, by the Indian Type Foundry, released under an open licence so there is no rights problem.
It is geometric, so its O is a true circle and its lowercase a is single storey. On a sequence built around a porthole and a rotating drum, the letters are made of the same shape as the subject. Bold at 118px with wide tracking for the title, Regular and SemiBold for credits, and all type added in After Effects rather than painted into the artwork.
There is a pleasing coincidence in the foundry, though I am wary of overclaiming it. Poppins ships with a Devanagari companion, while Urdu, which a British Pakistani family in 1985 would more likely read, uses Nastaliq. So the connection is nice rather than meaningful, and I would rather say that than pretend otherwise.
Fourteen frames, remade
With the style fixed, I rebuilt all fourteen storyboard frames using the same style string, changing only what is in the drum and letting the saturation climb across the sequence. You can see the frames below in this video sequence:
Held at arm’s length, you can see the temperature rise across the page. That is the clearest evidence I have that changing direction was the right call.
Frame 10 was the real test: Omar and Johnny forming out of the swirling cloth as two silhouettes, foreheads almost touching. No faces, no literalism. It carries more feeling than any of the grey versions managed. The faceting can do tenderness as well as politics, which is what I needed to know.
Sound, and a useful discovery
I also started thinking about the soundtrack, and realised my Week 3 audio plan had the same problem as the visuals: it was written for the grey version.
The structural idea survives. The laundrette generates the rhythm rather than music being laid underneath. What changes is the temperature.
I found two things worth having. The first is that 96 BPM makes everything line up: one bar is 2.5 seconds, ninety seconds is exactly 36 bars, and at 24 frames per second one bar is exactly 60 frames. Put a marker every 60 frames and every cut lands on a downbeat. (Thank you Google).
The second is the Pet Shop Boys. Their re-recorded “West End Girls” came out in October 1985 and hit number one within eight weeks, so it was genuinely the sound of the street this film is set on. They also wrote music for the 2019 stage adaptation of My Beautiful Laundrette. I am not sure whether I will use this but it provides a production brief: melancholy and danceable at the same time, which is the tone problem I have.
That sits alongside the Southall bhangra scene of the same moment, and the two together give me the collision the whole piece is about. Two kinds of British music from the same city in the same year, meeting inside a washing machine.
What I take from Week 4
Pre-production is still the creative work. But this week added something sharper. I am not 100% certain that I will be able to make each of the ‘assets’ this project requires and I have hedged on this for 48 hours before starting.
Research can tell you what is possible but it cannot tell you what you want. I built a careful, well-evidenced argument for a style and then found I did not want to look at the thing it produced, and no amount of further research would have told me that. Only making the image did.
“Second year is supposed to be about taking risks”. I had assumed that meant choosing something difficult. It turns out it also means being willing to bin a week’s reasoning when the evidence in front of your eyes disagrees with it.
Coming soon – the animatic! But in the meantime there are a lot of assets to make:
Back to front, as the worked example. Every other frame follows the same logic.
- Laundrette back wall, faceted planes
- Checkerboard floor
- Bench
- Machine body, orange casing
- Light wedge, the only “lighting” element, a flat painted shape whose opacity you animate
- Drum interior, dark
- Water spiral, square, larger than the porthole
- Flag shard
- Newspaper
- Terraced houses
- Whisky bottle and tumbler
- Banknotes
- Block-print panel
- Foam flecks
- Porthole bezel, chrome ring, sits in front of everything inside
- Glass highlight
- Type
Leave a Reply