Syed Abdul Wasay Ali · AI Creative Manager & AI Image Specialist

SYED ABDUL WASAY ALI, ai creative · image specialist. Text version of the interactive portfolio (images & video). Full site: https://syed-abdul-wasay-ali.github.io/syed-portfolio-video/

Building photorealistic product imagery, commercial visuals and production-ready AI workflows, real products placed into generated environments without losing product fidelity, composited with matched perspective, light and shadow. AI Creative Manager at Ogilvy · ex-AI Head Artist at cleanDirty.ai.

comfyui · flux · qwen · generative ai · product visualization

Contact: syedabdulwasayali@gmail.com · LinkedIn · GitHub · Resume (PDF)

About

I work at the intersection of creative direction, AI image generation and production workflows. My work focuses on using generative AI to create controlled, photorealistic visual assets for commercial applications, real products placed into generated environments without losing fidelity, edited and composited to a deliverable.

I'm from Hyderabad, India, with an art background. Before this I worked in customer service at Amazon, and in 2023 I resigned to pursue better opportunities and learn new things. I went all in on generative AI, and I haven't stopped learning new models since.

Most of my production happens in ComfyUI: image-generation workflows, reference-controlled generation, image editing, retouching and compositing, with LoRA training when a set needs consistency. I religiously check the AI subreddits, X accounts and Instagram pages that cover every new tool and breakthrough, so I know what's new the moment it lands, and I automate the workflows around it all with AI agents like Hermes, Codex and Claude Code.

I spend my days on the production side: building the pipelines, tuning the samplers, solving the problems that show up between an idea and a final file. Everything below is documented the way I work: real models, real parameters, real obstacles, and the ComfyUI graphs that got past them.

Experience

AI Creative Manager, Ogilvy

Mumbai, India. Apr 2025 - Present

AI-native production for client campaigns: image and video pipelines, product imagery systems, identity consistency, automation.

AI Head Artist, cleanDirty.ai

Delhi, India. (Remote) Sep 2023 - Feb 2025

Led AI art direction and production: custom LoRAs, image & video generation systems, mocap pipelines.

Skills

What I Solve

I build production-ready AI image systems for commercial product and lifestyle advertising, product accuracy, compositing, and repeatable workflows.

Where AI Meets Image Craft

A model gets you close. These are the controls that make the output a commercial deliverable.

Selected work, case studies (13)

Bedroom Lifestyle Comp

AI Creative Manager · Concept · 2026 · concept

Product/environment integration + perspective & light matching

Integrating a real product into an AI-generated bedroom while holding realistic scale, camera perspective, contact shadows, reflections and lighting direction, the environment changed, the product did not.

Spec: discipline: lifestyle + product integration · product: mattress + bedroom set (concept) · route: product → environment → perspective → light · lighting: morning window, camera-left · year: 2026

Contribution: bedroom art direction + light plan; environment generation, lens and framing matched; product placement + perspective match; light direction, temperature + contact shadows; final composite + grade

Stack: Own art direction + light plan, Environment generation, Midjourney v7, Reference-controlled placement, Perspective + scale match, Light direction, temperature + contact shadows, Grade + format crops

Reference-Controlled AI Imaging

AI Creative Manager · Concept · 2026 · concept

Reference images → controlled generation → production variations

Using reference imagery, structured prompting and AI image-editing workflows to maintain visual consistency while generating controlled variations, same product, six environments, zero drift.

Spec: discipline: controlled generation + variations · product: one product, six environments · route: reference → control → generation → edit · engine: midjourney v7 + chatgpt edit pass · year: 2026

Contribution: reference pack + prompt system; controlled generation across environments; identity-edit passes against the reference; per-frame clean-up + unified grade; variation set delivery

Stack: Structured prompt system (environment / camera / light), Reference conditioning, Midjourney v7, Edit passes, chatgpt relight + cleanup, Per-frame cleanup + unified grade, Variation set, six environments, one product

Media: minimal studio, variation 02 · warm luxury, variation 03 · daylight interior, variation 04 · wide, variation 05 · detail, knit + piping, same product · reference · prompt / control · generation · editing pass · variations

E-commerce Image System

AI Creative Manager · Concept · 2026 · concept

AI imagery designed for product pages, ads and mobile experiences

What happens to AI imagery after generation: an image set built for a product page (hero, lifestyle, detail, thumbnails) cropped and recomposed per slot so the page holds one product across desktop and mobile.

Spec: discipline: e-commerce imagery + storefront design · slots: banner · gallery · detail · mobile · ratios: 16:9 · 4:5 · 1:1 · deliverable: set + concept page (design study) · year: 2026

Contribution: storefront image set, hero, lifestyle, detail; slot planning, banner, gallery, thumbs; crops tuned for focal point + legibility; mobile recomposition; concept page build, desktop + phone

Stack: Storefront slot plan, banner, gallery, detail, mobile, Crops from one graded master set, Focal-point + legibility tuning, Mobile recomposition, Concept page, desktop + phone (design study)

Media: product pass, isolated master

Sleep brand image system · hero → lifestyle → macro → e-commerce

AI Creative Manager · Concept · 2026 · concept

A complete advertising surface for a sleep brand, built as one system: hero, bedroom lifestyle, night scene, macro comfort detail, unboxing, flatlay and a dark-studio banner with copy space. Twelve frames, one mattress, the same product in every image, extended from a repeatable core pipeline.

Spec: brand: sleep brand (concept) · product: mattress + bedding · deliverable: 12-frame image system · formats: 16:9 · 4:5 · 1:1 · engine: krea 2 turbo (fp8) · comfyui

Contribution: system design, one product, twelve frames; photoreal lifestyle set, morning, night, couple; studio + macro detail set; e-commerce frames, unboxing, flatlay, banner; edit passes + grade, one shoot, one look

Stack: Krea 2 turbo (fp8) · ComfyUI graph, Reference-image workflows, scene + product lock, Own GPU batch queue, full set from one pipeline, Edit passes, hands / packaging cleanup, Final grade + 16:9 / 4:5 / 1:1 crops

Media: hero, morning bedroom · studio, product profile · lifestyle, morning light · lifestyle, night · lifestyle, two sleepers · e-commerce, unboxing moment · detail, comfort press test · macro, knit weave · macro, comfort layers · macro, airflow structure · e-commerce, styled flatlay · e-commerce banner, dark studio, copy space

7up · concept ad (Seedance 2.5)

AI Creative Manager · Concept · 2026 · concept

My own original idea for a 7up concept ad: a colour-drained wasteland under a weathered 7up billboard that promises no more dull moments, one bottle, one first sip, and the whole world snaps into colour. Generated with Seedance 2.5.

Spec: brand: 7up · idea: own original concept · engine: seedance 2.5 · output: 1280×720 · 30s · year: 2026

Contribution: original concept + ad idea; seedance 2.5 generation; dull-world → bright-city turn

Stack: Own original concept + ad idea, Seedance 2.5 (text-to-video, closed-source), 1280×720 · 30s master

Media: 7up · concept film (30s) · the dull world · opening · the discovery · under the billboard · the bottle · in the dust · the first sip · the turn · the world goes bright · the city · 7up-bright future · sign-off · with the 7up mascot

Futuristic Neon Chase · MCP Blender env → Seedance 2.5

AI Creative Manager · Concept · 2026 · concept

A neon chase built in two passes: the environment, rider and camera path were made in Blender with the agent driving it over MCP and baked to a clean clay render, then Seedance 2.5 turned that pass into the final cinematic motion picture. The two videos sit side by side so the geometry work and the generated film can be judged against each other.

Spec: environment: blender · built via MCP · guide: clay render pass · 1280×720 · engine: seedance 2.5 · motion to video · output: 1280×720 · 30s · comparison: side-by-side master · 30s · year: 2026

Contribution: environment + camera built in Blender via MCP; clay render pass as the motion guide; seedance 2.5 motion-to-video render; side-by-side comparison master; timing + framing checks across both passes

Stack: Blender (driven over MCP), Clay render pass, Seedance 2.5 (motion-to-video, closed-source), Side-by-side comparison master

Media: side-by-side comparison · clay vs seedance 2.5 · blender clay pass · mcp environment · seedance 2.5 · final neon chase · neon chase · key frame · blender clay · key frame

Happydent White gum · concept ad (Seedance 2.5)

AI Creative Manager · Concept · 2026 · concept

My own concept and ad idea for Happydent White: a school annual-day performance where the lights go out and the smiles stay on. The imagery is generated with Seedance 2.5, and the film is directed and edited by me, down to the final packshot and its line: bright white. bright moments.

Spec: brand: happydent white · mint flavour · idea: own concept + ad idea · engine: seedance 2.5 · craft: directed + edited by me · output: 1920×1080 · 32s · year: 2026

Contribution: original concept + ad idea; direction; seedance 2.5 generation; edit + 32s cut; packshots + taglines

Stack: Own concept + ad idea, Seedance 2.5 (text-to-video, closed-source), Edit + 32s cut (own work), 1920×1080 master

Media: happydent white gum · concept film (32s) · the signature · a smile that lights the dark · the unwrap · happydent white packet · the audience · one evening at the show · the dark passage · mid-film · the stage · annual-day performance · end card · when the lights go out... keep your smile on · final packshot · bright white. bright moments.

History of Diamonds in India, timeline imagery (NDPF)

AI Creative Manager, Ogilvy · Ogilvy · 2026 · shipped

For the Natural Diamond Promotion Foundation I created the visual world of the "History of Diamonds in India" timeline: the Golconda diamond at twilight, diamonds resting on the old trade map, river panning for alluvial stones, plus the 1900 Surat workshop and 1950s-60s factory scenes, generated with closed-source image models so the timeline reads like real archival photography.

Spec: client: natural diamond promotion foundation · feature: history of diamonds in india timeline · engine: closed-source image models · route: era-correct historical imagery · year: 2026 · engagement: ogilvy

Contribution: historical photo prompting; closed-source image models; era look tuning; timeline asset delivery

Stack: Closed-source image models, Era-correct historical styling, Timeline asset production, Natural Diamond Promotion Foundation deliverable

Media: golconda era, diamond at twilight (key visual) · diamonds on the old trade map (key visual) · river panning for alluvial diamonds (key visual) · history of diamonds in india, timeline section (ndpf.in) · history of diamonds, first discovery & ancient trade (ndpf.in) · history of diamonds, golconda era & 1665 (ndpf.in)

Campaign: see the timeline live, ndpf.in

IFB Christmas 2025 · the Santa finger snap (Wan Animate)

AI Creative Manager, Ogilvy · Ogilvy · 2025 · shipped

The IFB Christmas film "Because some miracles do happen overnight" is built around one tiny gesture: a finger snap. It was also the hardest shot to make. The top closed-source video models of the time, Google Veo 3 and Kling 2 among them, could not generate a natural snap. So I recorded my own hand performing the snap and drove it through Wan Animate with a Santa hand reference image, until the gesture read real. The side-by-side shows the hand capture against the generated Santa hand.

Spec: client: ifb appliances · film: because some miracles do happen overnight · moment: the santa finger snap · engine: wan animate · open-source · motion source: own hand capture · failed first: veo 3 · kling 2 · engagement: ogilvy

Contribution: wan animate motion transfer; own hand capture as the motion source; santa hand reference setup; finger snap generation + iteration; closed-source model tests (veo 3 · kling 2)

Stack: Wan Animate (open-source), Own hand capture (motion source), Santa hand reference image, Finger snap generation, Open-source-first pipeline, IFB Christmas 2025 film deliverable

Media: ifb christmas 2025 · "because some miracles do happen overnight" (youtube) · finger snap · hand capture vs wan animate (side by side) · hand capture · full clip · wan animate · full clip

Campaign: watch the film, ifb christmas 2025 (youtube)

Vi 5G FanFest "Cricky" · clay character LoRA (IPL 2025)

AI Creative Manager, Ogilvy · Ogilvy · 2025 · shipped

Vi 5G FanFest turned IPL 2025 into a match-day cricket quiz across Instagram, X and Facebook, with vouchers worth ₹5,000 up for grabs. The face of the campaign was one clay-style character, and every card had to show that same character. I trained that character as a LoRA on Flux, and the look held for the whole season.

Spec: client: vi (vodafone idea) · campaign: vi 5g fanfest · cricky quiz · base: flux · lora: clay-style character · own training · channels: instagram · x · facebook · year: 2025 · engagement: ogilvy

Contribution: clay-style character lora training · flux; character consistency across the season; match-day quiz card generation; one look held across instagram · x · facebook

Stack: Vi 5G FanFest · match-day quiz campaign, Character LoRA training (Flux base), Clay-style character consistency, Match-day quiz card generation, Delivered on instagram · x · facebook

Media: fanfest quiz q4 · del vs hydr · 30 mar · fanfest quiz q4 · lkn vs guj · 12 apr · fanfest quiz q1 · del vs mum · 13 apr · fanfest quiz q4 · mum vs hyd · 17 apr · fanfest quiz q2 · blr vs raj · 24 apr · fanfest quiz q5 · blr vs che · 03 may

Campaign: the cricky challenge launch reel (instagram)

Lacta Christmas AI app, photo to avatar "Compartilhe o seu Natal"

AI Creative Manager, Ogilvy · Ogilvy · 2024 · shipped

For Lacta’s Christmas campaign I created the images that drive the AI app: a consumer uploads a photo and it comes back as an AI avatar inside a personalised Christmas story, ready to share. The campaign itself ran as an Instagram reel and the app segment was the heart of it.

Spec: client: lacta (mondelez) · campaign: compartilhe o seu natal · feature: photo → ai avatar + christmas story · engine: ai image workflow · season: christmas 2024 · year: 2024 · engagement: ogilvy

Contribution: ai image generation; photo → ai avatar conversion; identity-safe renders across faces; seasonal look lock; batch iteration & cleanup

Stack: AI image generation, Avatar + identity workflow, Seasonal style consistency, Lacta Christmas campaign, Instagram reel delivery

Media: lacta christmas app, campaign reel (instagram) · ai avatar frame, christmas moment

Campaign: view the campaign reel, instagram

Cadbury Celebrations, "Creating Memories Never Clicked"

AI Creative Manager, Ogilvy · Ogilvy · 2024 · shipped

The film is about the memories that were never clicked: the grown-up characters in the Cadbury Celebrations campaign were de-aged back to their younger selves. I built the image workflow in ComfyUI that did the de-aging, using SDXL checkpoints with a de-aged LoRA plus face swap workflows.

Spec: client: cadbury celebrations · base: sdxl (open-source) · lora: de-aged lora · route: face swap workflows · engine: comfyui image workflow · film: 60s · hindi · year: 2024 · engagement: ogilvy

Contribution: ai image generation; de-aged lora training; face swap workflows; identity drift checks; shot iteration & refinement; comfyui workflow development

Stack: ComfyUI, SDXL checkpoints (open-source), De-aged LoRA, Face swap workflows, Image de-age pipeline, Own local GPU, Cadbury Celebrations deliverable

Media: cadbury celebrations, creating memories never clicked (60s · youtube) · de-aged frame, schoolyard scene

Campaign: watch the public release, youtube

Cadbury "The Story of Us" (Disney+ Hotstar · directed by Zoya Akhtar)

AI Creative Manager, Ogilvy · Ogilvy · 2024 · shipped

An animated brand film built frame-by-frame from open-source models: style LoRAs on the SD 1.5 stack, trained and re-trained until the story held. No Nano Banana. No Seedance. Just the GPU and the graph. The Zoya Akhtar-directed cuts also turned real couples into Disney-style 3D cartoon characters through a ControlNet-guided app, telecast on Disney+ Hotstar.

Spec: client: disney+ hotstar (now jiohotstar) · director: zoya akhtar · base: sd 1.5 (open-source) · lora: style lora · own GPU · guide: controlnet image → disney 3d · year: 2024 · engagement: ogilvy

Contribution: style lora training, sd 1.5 stack; character consistency across cuts; controlnet-guided conversion app; shot iteration & refinement; identity locks frame by frame

Stack: SD 1.5 (open-source), Style LoRA training, ControlNet (composition + gesture guide), Image → Disney-style 3D conversion app, Own local GPU, Frame sweeps / parameter tuning, Sequence assembly, Disney+ Hotstar deliverable

Media: final film, the story of us (hotstar promo) · final film, "now" cut (1080p master) · story of us, valentine reel (sd 1.5 · controlnet) · story of us, valentine still 01 · story of us, valentine still 02 · story of us, valentine still 03 · story of us, valentine still 04 · story of us, valentine still 05 · story of us, valentine still 06 · story of us, valentine still 07 · story of us, valentine still 08 · story of us, valentine still 09

Concept images, AI × e-commerce

AI Product Photoshoot Images

Character consistency

The same face from the first frame to the last: character sheets built before a shot, character LoRAs trained in-house, de-aging and conversion workflows that keep the identity intact.

Concept ads & personal work

Spec work, campaign concepts and experiments, AI-made, no client brief required. Click any piece to inspect.

ComfyUI workflows

21 brands I've worked within Ogilvy

Click a logo with media for the stills, animatics and final films. 21 brands and counting.

Contact

Available for AI image generation, product visualization, commercial creative and AI production workflow projects.

syedabdulwasayali@gmail.com · linkedin · GitHub · media/resume/Syed-Abdul-Wasay-Ali-Resume.pdf

based in hyderabad, india · remote-ready