Aller au contenu

Image Pipeline — From Phone to Publication

Goal: Practice teams shoot with phones. Central team processes with AI tools. Every practice gets professional, editorially consistent images without hiring a photographer for each site.

Overview

Practice team                    Central team                      Aletheia
────────────                     ────────────                      ────────
1. Capture (phone)          →    2. AI Processing (batch)     →    3. Upload
   - Follow shot list            - Background swap                 - Alt text
   - Basic framing rules         - Light/color correction          - Auto WebP + srcset
   - Raw photos (no filters)     - Consistent crop                 - LQIP blur
                                 - Upscale if needed
                                 - QA review

Step 1 — Capture (Practice Team)

Equipment

A recent smartphone is sufficient. No camera gear needed.

Setting Value Why
Camera Rear camera (never front/selfie) Higher resolution, better lens
Mode Portrait mode OFF for rooms, ON for people Portrait mode AI can distort edges
Flash OFF always Flash creates harsh shadows and blown-out faces
HDR ON Balances bright windows and dark interiors
Format HEIF/JPEG at max resolution Don't shoot in RAW (too large, unnecessary)
Grid ON (rule of thirds) Helps with framing
Filters NONE Central team handles color grading

Shot Lists

Send the relevant shot list to each practice. They check off items and send raw photos via a shared drive (Google Drive, or direct Aletheia upload).

Team Portraits — Shot List

For each team member, capture 3 shots (central team picks the best):

  1. Head and shoulders — Subject centered, slight angle, smiling, eyes to camera
  2. Slightly wider — Chest up, arms relaxed or crossed
  3. Backup — Same as #1 with different expression

Setup instructions for the practice: - Find a spot with a plain, light-colored wall (white, light gray, beige) - Stand 1.5m from the wall (avoids shadow on wall) - Photographer stands 2m from subject - Natural light from a window to one side (45° angle) — best between 9-11am or 2-4pm - No overhead fluorescent lights if possible (they add green cast) - If multiple windows, use the largest one - Subject wears white coat (blouse blanche), closed - Take all team photos in the same session, same spot, same day

Common mistakes to flag: - ❌ Bathroom mirror selfies - ❌ Cropped group photos - ❌ Different background per person - ❌ Subject standing against the wall (creates shadow) - ❌ Backlit (window behind subject) - ❌ Instagram filters applied

Cabinet Interior — Shot List

Shot Framing Notes
Reception / waiting room Wide angle, landscape, from entrance Tidy up: fresh magazines, no coats, no personal items
Treatment room (main) Wide angle, landscape, from doorway Chair centered, screen on, lights on, no patient
Treatment room (detail) Medium shot of chair + equipment Show technology context
Building exterior Landscape, signage visible Daytime, good weather, no cars blocking
Corridor / hallway Landscape, end-to-end Shows space, accessibility

Setup instructions: - Shoot during off-hours (no patients, no clutter) - All lights ON (even if natural light is sufficient — avoids dark corners) - Clean everything: screens, chairs, counters, floors - Remove personal items, food, jackets, bags - Close cabinet doors and drawers - Landscape orientation only (16:9 crop in pipeline)

Equipment Close-ups — Shot List

One photo per notable piece of technology: - Scanner 3D / Cone Beam - Scanner intra-oral - Panoramique numérique - Autoclave / sterilization area (for charte qualité page) - Any distinctive equipment (microscope, laser, CFAO)

Framing: Medium shot showing the equipment in context (mounted, connected, in the room). Not a product photo — show it in its environment.

Step 2 — AI Processing (Central Team)

Tool Stack

Task Recommended tool Alternative Cost
Background removal/swap Photoroom ClipDrop, remove.bg Photoroom Pro ~10€/mo
Light & color correction Adobe Lightroom (with presets) Luminar Neo, Topaz Photo AI Lightroom ~12€/mo
Upscaling Topaz Gigapixel AI Real-ESRGAN (free, open source) Topaz ~100€ one-time
Batch processing Lightroom (presets + batch export) ImageMagick + scripts —
Background generation Photoroom AI backgrounds Adobe Firefly Included in Photoroom Pro
Face enhancement Topaz Photo AI (face recovery) — Included in Topaz bundle

Total cost: ~25€/month for the full pipeline. Compares to ~400-500€ per practice for a photographer.

Processing Pipeline — Team Portraits

Raw photo (phone)
    │
    ├─ 1. Background removal (Photoroom / ClipDrop)
    │     → Transparent PNG
    │
    ├─ 2. Background replacement
    │     → Consistent background per practice:
    │       Option A: Solid neutral (light gray #E8E8E8)
    │       Option B: Soft gradient (white to light gray)
    │       Option C: Blurred practice interior (same for all members)
    │       Option D: AI-generated clinical environment
    │
    ├─ 3. Light & color correction (Lightroom preset)
    │     → White balance normalized
    │     → Exposure balanced
    │     → Skin tone consistency
    │     → Subtle warmth (+5 temp, +5 tint)
    │
    ├─ 4. Crop to 3:4 portrait
    │     → Head at ~20% from top
    │     → Shoulders visible
    │     → Centered
    │
    ├─ 5. Upscale if needed (Topaz)
    │     → Minimum 600x800px output
    │     → Face enhancement ON (sharpens without over-smoothing)
    │
    └─ 6. Export
          → JPEG quality 92, sRGB
          → Filename: {practice-code}_{lastname}_{firstname}.jpg
          → Upload to Aletheia with alt text

Processing Pipeline — Cabinet Interiors

Raw photo (phone)
    │
    ├─ 1. Perspective correction (Lightroom / auto)
    │     → Straighten verticals (walls, door frames)
    │     → Level horizon
    │
    ├─ 2. Light & color correction (Lightroom preset)
    │     → "Helios Interior" preset:
    │       - Shadows +30 (open up dark corners)
    │       - Highlights -20 (recover bright windows)
    │       - White balance: ~5500K (neutral daylight)
    │       - Vibrance +10 (subtle color boost)
    │       - Clarity +15 (detail without harshness)
    │
    ├─ 3. Object removal if needed (Lightroom AI / Photoshop)
    │     → Remove distracting items missed during capture
    │     → Remove reflections in screens/glass
    │
    ├─ 4. Crop to 16:9
    │     → Landscape orientation
    │     → Subject centered or rule-of-thirds
    │
    ├─ 5. Upscale if needed
    │     → Minimum 1920x1080px for heroes
    │     → Minimum 800x500px for cards
    │
    └─ 6. Export
          → JPEG quality 92, sRGB
          → Filename: {practice-code}_{room-type}.jpg

Processing Pipeline — Equipment Close-ups

Same as cabinet interiors, with one addition: - After color correction, apply a subtle vignette to draw focus to the equipment - Crop to 16:10 (card format) or 1:1 (if used in a grid)

Step 3 — Editorial Consistency

The "Helios Look"

All processed images should share these characteristics:

Attribute Target How
Temperature Warm neutral (not cold/clinical, not orange) Lightroom: ~5500K, tint +3
Contrast Medium — not flat, not punchy Lightroom: contrast +10, blacks -5
Saturation Natural — slightly muted, not vivid Lightroom: vibrance +10, saturation -5
Shadows Open — no crushed blacks, dark corners brightened Lightroom: shadows +20-30
Sharpness Crisp but not over-sharpened Lightroom: amount 40, radius 1.0
Skin tones Natural, consistent across team members Calibrate per session using HSL panel

Create a single Lightroom preset ("Helios Portrait" + "Helios Interior") and apply to all photos. Adjust exposure per-image, but keep the look consistent.

Background Consistency Rules

Photo type Background approach
Team portraits (same practice) All members get the exact same background
Team portraits (across practices) Backgrounds can differ per practice, but same style (e.g., all solid, all gradient)
Cabinet interiors Real photos — no background replacement
Hero images Real photos preferred. AI composite only if no usable hero shot exists

What AI Should NOT Do

  • ❌ Generate faces — All team photos must be real people. AI-generated faces are detectable, unethical, and would violate patient trust.
  • ❌ Alter body/face shape — Enhance lighting and color, never alter anatomy.
  • ❌ Create fake equipment — Equipment photos must show actual equipment at the practice.
  • ❌ Fabricate clinical results — Before/after photos must be real clinical photography. Never AI-enhanced or generated.
  • ❌ Over-smooth skin — Face enhancement should sharpen and correct lighting, not create a plastic look. Keep pores, lines, natural texture.

What AI CAN Do

  • ✅ Remove/replace plain backgrounds on portraits
  • ✅ Correct white balance, exposure, color cast
  • ✅ Remove distracting objects (a jacket on a chair, a stray cable)
  • ✅ Straighten perspectives on interior shots
  • ✅ Upscale phone photos to publication resolution
  • ✅ Generate neutral/professional backgrounds for portraits
  • ✅ Batch-process for consistency (same preset on all photos from one practice)

Workflow Per Practice

Timeline

Day Task Who
D1 Send shot lists + capture guide to practice Central team
D2-D5 Practice takes photos, uploads to shared drive Practice team
D6 Central team reviews raw photos, requests retakes if needed Central team
D7-D8 AI processing pipeline (batch) Central team
D9 QA review: consistency, quality, crop, resolution Central team
D10 Upload to Aletheia with alt text Central team

Total: ~10 days from request to publication-ready images. Can run in parallel with content creation.

Quality Checklist Before Upload

  • [ ] All team portraits have the same background
  • [ ] All team portraits have consistent color temperature
  • [ ] All interiors are perspective-corrected and level
  • [ ] All images meet minimum resolution for their usage
  • [ ] No AI artifacts visible (check edges of removed backgrounds, faces)
  • [ ] Filenames follow convention: {practice-code}_{subject}.jpg
  • [ ] Alt text prepared for each image
  • [ ] No patient visible in any photo
  • [ ] No personal items visible (names on screens, documents, bags)

Cost Comparison

Approach Per practice Quality Consistency
Professional photographer 400-600€ Excellent Good (varies by photographer)
Phone + AI pipeline 25€/mo (tools) Very good Excellent (batch presets)
Phone without processing 0€ Poor to mediocre Poor (varies wildly)

The AI pipeline pays for itself after the first practice. By practice #3, it's saving thousands compared to hiring photographers in each city.

Before/After Case Studies — Exception

Clinical before/after photos cannot go through the AI pipeline. They must remain unprocessed (except basic crop and white balance) to maintain clinical integrity. See image-guidelines.md → Before/After section for requirements.

The only acceptable processing: - Crop to 4:3 - White balance correction (neutral, no warming) - Anonymization (blur/crop to exclude identifying features if not already intraoral)