Home
About
Contact
All projects
AI image processing

Vivien Vance - AI Photo Enhancement

Professional photo enhancement with custom style personas

An AI-powered image enhancement system that works with persona-specific styles. The client uploads the photo, the system enhances, retouches, and optimizes it according to the selected style.

3x
faster processing
5+
custom personas
0
manual retouching

The challenge#

Vivien Vance is a professional photography studio specializing in weddings, business portraits, fashion photography, and family sessions. The studio produces thousands of images annually, and behind every shot lies significant post-production work. The shoot itself typically takes a few hours, but post-processing images often takes several times longer than the on-location work. This invisible work is unknown to most clients, but for the photographer, it's the most time-consuming and repetitive part of daily routine.

Each client segment requires a different style: fashion photos need strong contrast and vibrant colors, business portraits call for a restrained, natural look, and wedding photos demand warm, romantic tones. This isn't a simple filter swap. The differences between styles range from white balance adjustments to the degree of skin retouching, from background handling to sharpness fine-tuning. An experienced retoucher needs to keep every client's expectations in mind and apply them manually to every single image.

An experienced retoucher spends 10-15 minutes on a single image: exposure correction, white balance, skin retouching, eye color enhancement, background handling. Post-processing a 200-image wedding series means full working days. Clients, however, expect a fast turnaround: a newly married couple doesn't want to wait weeks for their photos, and fashion agencies need final materials within days. This pressure forced the studio to either hire employees exclusively for retouching or turn down orders due to capacity constraints.

Consistency was also a problem. The same client's images were processed over multiple days, and due to fatigue, changing light conditions, and human subjectivity, the style of the last images differed from the first ones. In a wedding album, this is especially noticeable: if the first thirty images have warm tones but the retoucher instinctively shifts toward cooler tones by the end of the series, the overall effect isn't unified. The client can't articulate the problem, but they feel something is off.

The goal: build a system that automates the repetitive, mechanical part of post-production while preserving the unique style. We didn't want to replace the photographer, but to speed up their work so they can focus on creative and client-facing tasks.

iAI photo enhancement vs. traditional batch processing

Photoshop and Lightroom batch processing is rule-based: it applies the same settings to every image (e.g., +0.5 exposure, -10 vibrance). If the image is too dark or too bright, the rule doesn't adapt. A dimly lit church wedding shot and a bright outdoor portrait receive the same correction, which improves one but worsens the other. The photographer ultimately has to manually override the batch settings, which defeats the very time savings batch processing promised.

AI photo enhancement is different: the model understands the image content and makes decisions accordingly. If the portrait background is dark, it doesn't brighten unnecessarily. If the skin tone is already good, it doesn't over-retouch. It makes unique decisions for each image, but the overall result stays consistent because the persona defines the target. The persona essentially tells the AI: "This is the result I want to achieve," not "Apply these three technical settings." The model itself decides how to reach the expected result, adapting to each image's characteristics.

This is the difference: batch processing does the same thing to every image; AI photo enhancement achieves the same result on every image. Just as an experienced retoucher handles different lighting conditions differently, but the result remains stylistically consistent, the AI also makes unique decisions per image while the end result stays unified.

The solution in detail#

1

Persona setup — defining the style

The soul of the system is the persona system. Each client (or style category) has a "persona" that describes the expected end result in detail. The persona doesn't contain technical parameters, but a style description in the photographer's own language, which the AI interprets and applies:

  • Color palette: Warm/cool tones, saturation level, contrast
  • Retouching level: Natural (minimal) / Medium / Heavy (magazine quality)
  • Special instructions: "Eye color enhancement," "Background blur," "Skin smoothing with natural texture"
  • Export settings: Resolution, format (JPEG/PNG/TIFF), quality, watermark

The photographer sets up the persona once, then it's automatically applied to every image. When a returning client comes for another session, the previously used persona can be recalled instantly, guaranteeing style continuity with previous images.

2

Upload — simple, bulk

The photographer uploads raw images through the n8n web interface or simply copies them to a designated Cloudflare folder. The system watches the folder and automatically sends every new image for processing. This means the photographer can swap cards on location, transfer the raw files to their computer, drag them into the designated folder, and processing begins immediately.

Supported formats: JPEG, PNG, TIFF, and raw RAW files (CR2, NEF, ARW). For RAW files, the system first performs a default development, then applies the persona settings. During upload, the system automatically recognizes the format, performs the necessary conversion, and notifies the uploading photographer if a file is corrupted or in an unsupported format.

3

AI processing — content-based image enhancement

The OpenAI Vision API analyzes the image and performs enhancements based on the persona instructions. Processing doesn't happen in a single step: the system first assesses the image characteristics (lighting conditions, number and position of faces, background type), then prioritizes the necessary interventions accordingly:

  • Exposure and white balance correction based on image content
  • Skin retouching at the specified level (preserving natural texture)
  • Color grading aligned to the persona's color palette
  • Sharpness and detail optimization

We tuned the AI to not over-retouch: the goal is to enhance natural appearance, not create a "plastic" effect. The system errs on the side of less intervention rather than too much, because slight under-retouching is far less jarring than excessive smoothing.

4

Quality control — the automatic safety net

The system compares every enhanced image against the original and performs multi-level checks. This step ensures no image leaves the system below the expected quality, and the photographer can use the automated process with confidence:

  • If the change is too drastic (e.g., the face became unrecognizable), it stops processing and flags it for the photographer
  • If the image is still too dark or too bright after enhancement, it retries with different settings
  • The photographer can review results in a gallery and regenerate any image with different persona settings in a single click

The system learns from the photographer's feedback: if an image is rejected or regenerated, that information helps fine-tune the persona for future processing.

Before and after#

Előtte
  • 10-15 minutes of manual retouching per image
  • 200-image series: 2-3 full working days
  • Inconsistent style due to fatigue and subjectivity
  • Manual settings in Lightroom for every client
  • RAW processing + retouching as separate steps
Utána
  • Seconds: AI processing per image
  • 200 images: < 1 hour automatic processing
  • Consistent quality, persona guarantees the style
  • Persona set once, automatically applied
  • Upload to export in a single automatic process
vivienvance-ai.makeden.hu/personas

Persona management — Style presets

PersonaTypeRetouchingColor paletteImages processed
Wedding RomanticWeddingMediumWarm, pastel847
Business PortraitPortraitNaturalNeutral, sharp234
Fashion EditorialFashionHeavyVibrant, high-contrast156
Family NaturalFamilyLightWarm, natural412
E-commerce ProductProduct photoMinimalWhite background, clean89

Each persona contains a detailed description that the photographer can modify at any time.

Results in numbers#

MetricBeforeAfter
Processing time/image10-15 minSeconds (AI)
200-image series2-3 working days< 1 hour
Style consistencyVariablePersona-guaranteed
Monthly capacity~500 images2,000+ images
Manual retouching neededEvery imageOnly special cases (~5%)

The studio can process three times as many images in the same amount of time. The photographer focuses on creating and client relationships, not repetitive post-production. Consistency has improved because the persona system guarantees every image is processed in the same style. The freed-up time has enabled the studio to introduce new services, such as express 24-hour turnaround for wedding photos — something previously physically impossible.

How to apply this in your business#

Not just for photographers — business applications of AI image processing

AI image processing is applicable across numerous business domains, and the persona-based approach tested in the Vivien Vance project is adaptable to virtually any industry. The key is the same everywhere: you define the desired end result, and the system achieves it consistently, regardless of input image quality.

  • E-commerce: Automatic background removal, resizing, and optimization of product photos for webshops. Standardizing images across a 500-product catalog can be done in hours instead of days. Especially useful during seasonal campaigns when large volumes of product photos need rapid processing.
  • Real estate: Automatic enhancement of property photos (lighting, colors, HDR effect). A real estate agency photographs 10-20 properties daily, with 15-20 photos needed per listing. AI image processing ensures dark rooms still look appealing and every listing maintains consistent visual quality.
  • Marketing: Batch optimization of social media images for different platforms (Instagram, LinkedIn, Facebook). Automatic cropping, resizing, and style optimization of a single source image per platform, with brand-matching filters.
  • Documentation: Improving scan quality of documents before OCR processing. Improved image quality can increase OCR accuracy by 30-40%, which directly means less manual correction.

The key: define precisely what you want (persona), and the system will consistently deliver it. You don't need to be an AI expert. Setup happens in your professional language (color palette, mood, style, expected quality), not in technical parameters. The system adapts to you, not the other way around.

If you'd like AI-powered image processing for your business, book a free consultation.

Tech stack#

ToolRole
n8nUpload monitoring, processing pipeline, quality control
OpenAI APIImage analysis and AI-powered enhancement (Vision API)
CloudflareImage storage and web interface hosting
AirtablePersona management, processing log, client registry
Technologies used
n8n
OpenAI API
Cloudflare
Airtable