Photography & Imaging
Multi-Angle Campaign Reshoot
1 product photo → one campaign, held at precise product fidelity, iterated to every ratio and format its venues need — online and offline.
Best for: DTC brands, ecommerce agencies, fashion brands, out-of-home and print campaigns
Send one product or on-model photo. Get the campaign back as finished, venue-ready creative at every ratio the media plan runs — 16:9 for airport and billboard digital, portrait 2:3 for magazine and lobby screens, 9:16 for transit and Stories, 4:5 for feed and the PDP — with the product itself held to spec across all of them. A vision model reads the product first, then writes a custom prompt for each framing. Extreme framings run through a built-in critic-refiner loop: a second vision model scores each output for product fidelity, scale, proportions and perspective accuracy, rewrites the prompt to fix what's wrong, then regenerates. You touch nothing until every frame passes.
What you provide
- 1 product or fashion image (any angle)
What you get back
- One campaign concept, finished at every venue ratio — 16:9 airport / billboard · 2:3 magazine / lobby screen · 9:16 transit / Stories · 4:5 feed / PDP
- Precise product fidelity held across every framing and format
- Extreme framings refined through an automatic critic-refiner pass
Real outputs
- Airport Digital
- Highway Billboard
- Magazine — Full Page
- Lobby Screen
- Transit Poster
- Story / Reels
- Paid Social — Feed
- PDP / Catalogue
Live at every format
How it works
1 product or fashion image (any angle)
Your brand rules, palette, product facts, and approved references — applied to every prompt in the run.
- 01Garment Analysis
- 02Per-Angle Prompt Generation
- 03Parallel Generation: Run 1
- 04Critic Loop
- 05Prompt Refinement
- 06Refined Generation: Run 2
Flux Kontext Pro · GPT-4o Vision
Automated evaluation pass against the brief.
Final output requires human approval before delivery.
One campaign concept, finished at every venue ratio — 16:9 airport / billboard · 2:3 magazine / lobby screen · 9:16 transit / Stories · 4:5 feed / PDP
The bounded path every run follows — your inputs in, the real delivered output on the right. Step detail below.
- 01
Garment Analysis
A vision LLM reads the source image and writes a detailed prose analysis of the garment: fabric, fit, silhouette, colour, and movement characteristics. Output is a structured JSON description used to seed every downstream prompt.
- 02
Per-Angle Prompt Generation
A photography director LLM reads the garment analysis and writes a bespoke prompt for each camera framing, with precise physical camera placement, focal points, and lighting matched to the source image.
- 03
Parallel Generation: Run 1
Every framing is generated simultaneously using Flux Kontext Pro, with the original product image as reference. The framings that don't need correction are delivered as final outputs after this step.
- 04
Critic Loop
The extreme framings — overhead, low, and wide — each pass through a GPT-4o Vision critic that scores the output against 5 realism criteria: character scale, spatial relationships, body proportions, perspective accuracy, and spatial impossibilities.
- 05
Prompt Refinement
A specialised LLM for each framing reads the critic's analysis alongside the original prompt and rewrites it to fix the identified issues, correcting scale, camera height, and perspective while preserving the exact subject, clothing, and environment.
- 06
Refined Generation: Run 2
The flagged framings are regenerated using their corrected prompts, passing the scale and realism bar the first pass often misses. They are delivered alongside the final outputs from Run 1.
Brand, factual, and quality controls
- Automated evaluation pass against the brief.
- Human quality review before delivery.
What this workflow does not automate
- Requires approved product or brand source assets as input. It does not originate new photography or footage from scratch.
What you own
- Provide source product images or footage.
- Review and approve outputs before publish.
Typical turnaround
Illustrative ranges based on comparable production work, not a guaranteed delivery time. Actual turnaround is scoped per engagement.
Operating evidence
Verified operating evidence for this workflow is not yet published. Outputs shown above are real production results.
Related workflows
- Photography & Imaging
Editorial Camera-Angle Set
1 model image → 6 cinematic camera angles for the editorial. No second shoot.
- Photography & Imaging
On-Brand Static Ad Creatives (Packshots)
Drop a URL → 3 on-brand packshots. The AI reads your brand first.
- Photography & Imaging
Motion 360° Spin Product Video
4 product images → a smooth, loopable 360° spin video. No turntable rig.
Part of the Creative Production System.
Frequently asked questions
Why do only the extreme framings go through the critic loop?
Straight-on framings — front and three-quarter — don't suffer from the scale and perspective distortion that affects low, wide, and overhead angles. Running them through the loop would add time with no quality gain.
What does the critic actually check?
Character size relative to the environment, shadow and contact-point realism, head-to-body ratios, perspective line accuracy, and any physically impossible spatial relationships — all rated on a 1–10 realism scale with location-specific feedback.
Does it work with reflective or transparent products?
Yes. The garment analysis step handles material characteristics including glass, chrome, and translucent surfaces, which are passed into every generation prompt.
How many total AI calls does one run consume?
16 nodes: 1 I2T analysis, 1 LLM prompt generator, 6 parallel I2I generations, 4 I2T critics, 4 LLM correctors, 4 refined I2I generations. Approximately 48 Pletor credits per full run.
Prove this workflow on one bottleneck
Bring one recurring production task and your approximate monthly volume. You leave with a fit decision, not a custom architecture session.
Book a strategy call