Tom Ford Beauty approached Bajibot to create an AR filter for the Ombré Leather fragrance campaign — one of the brand's signature masculine-yet-sensual scents known for its raw leather and tobacco notes. The challenge was translating a deeply tactile, luxurious scent identity into a purely visual AR experience that would feel native to Instagram, premium enough to represent the Tom Ford brand, and technically flawless on the constrained canvas of a mobile camera filter.
AR filters for luxury brands demand a different bar: every frame must feel editorial, not gimmicky. The filter had to work across diverse skin tones and face shapes, load within Instagram's size limits, and perform at 30fps+ on mid-range phones — while evoking the leathery warmth of the fragrance without any audio or scent cues.
Before a single pixel was placed, I mapped every technical and brand constraint. Instagram AR filters impose a strict 4MB limit, 30fps floor, and face-tracking APIs that behave differently across iOS and Android. Simultaneously, Tom Ford's brand guidelines required a specific palette — near-blacks, warm ambers, off-whites — and zero use of the logo in motion-tracked contexts. I created a constraint matrix pairing each brand requirement against its technical feasibility in 8th Wall, flagging conflicts early before creative direction was locked.
Ombré Leather's identity is tactile — the campaign imagery is all shadow, texture, and smoke. My challenge was translating that sensory world into light-based AR. I developed a 'leather wash' concept: a face-conforming overlay using a custom normal-mapped leather texture that catches virtual light dynamically, combined with a warm-toned particle haze evoking tobacco and amber. I presented three concept directions to Vincent and the client: Smoke Veil (atmospheric), Leather Mask (structural), and Golden Hour (warm, soft). The client selected Leather Mask as the hero with Golden Hour color grading applied to the background feed.
I built the filter in 8th Wall's WebAR engine using A-Frame components with custom GLSL shaders for the leather surface effect. Key technical decisions: used 8th Wall's face mesh (not Spark AR) for more precise contour tracking across diverse face shapes; implemented a real-time ambient light estimation API to adjust the specular highlight intensity based on the user's environment lighting; and built a fallback for older Android devices that dropped the specular layer to maintain framerate.
I ran a structured QA matrix across 12 device/OS combinations — iPhone 12 Pro, Samsung Galaxy S21, Pixel 5, and older mid-range Androids. Tracked framerate, texture rendering fidelity, and face-tracking accuracy. Caught two issues: leather texture z-fighting on wider face shapes (fixed via mesh offset adjustment), and oversaturation of the amber grade on OLED screens (corrected with display-adaptive tone curve). Submitted final build to Tom Ford Beauty for brand approval; received one round of color temperature feedback (slightly cooler highlights) and shipped.
Spark AR had tighter Instagram integration but inferior face mesh fidelity on non-iPhone devices (~60% of Instagram users). 8th Wall provided more consistent contour tracking across Android hardware. Accepted slightly higher integration complexity for inclusive tracking quality. Result: consistent face tracking across all 12 tested device/OS combos — no significant tracking failures post-launch.
A static leather texture looks flat under varied real-world lighting. I implemented ambient light estimation API to dynamically adjust specular highlight intensity — so the filter looked rich in dim rooms and equally polished outdoors. Added ~3 days of development but dramatically elevated the premium feel. Result: press reviewers specifically noted the 'exceptionally polished' visual quality in Vogue and WWD.
Smoke Veil obscured the face too heavily for social sharing; Golden Hour was too generic for a signature scent. Leather Mask directly referenced the fragrance's core material identity and created a distinctive, shareable look that communicated the brand without logo placement. Result: client selected unanimously. The structural face mapping became the visual signature cited in all press coverage.
This project taught me that luxury brand AR is fundamentally a material translation problem — not a tech showcase. The filter succeeded because I started from the fragrance's sensory identity (leather, smoke, amber warmth) and worked backward to the shader, not the reverse. The constraint matrix I built at the start proved its value when a potential logo-in-motion conflict was caught in week one rather than week eight of development.
If I were doing this again, I'd push harder to test with a broader set of real users before brand review — the z-fighting issue on wider face shapes only surfaced because I specifically sought out diverse test participants. That kind of inclusive testing should be a standard first step, not a late QA pass.