Creative Analysis & 8 Pillars

Deep dive into the AI deconstruction engine, 8 Pillars of Creative Taxonomy, Demucs audio separation, and Timeline DNA.

AdHelix transforms monolithic video files and static image creatives into structured, queryable data entities. Every creative is decomposed into quantifiable visual, auditory, temporal, and narrative attributes.

1. Triggering an Analysis

Analysis is handled as an asynchronous queued background job:

Open Creative Library (/creatives) or Queue & Processing (/analysis).
Select an unanalyzed synced creative or upload a local .mp4, .mov, .webp, or .png file.
Click Analyze Creative.
The job is queued in the multi-modal pipeline (internal/analysisorch/), which extracts keyframes, separates audio tracks, transcribes speech, and grounds detected entities.
Upon completion, click the completed card to open the Analysis Detail Studio (/analysis/[creativeId]).

2. The 8 Pillars of Creative Taxonomy

AdHelix standardizes creative attributes into 8 canonical taxonomy pillars:

mindmap
  root((8 Pillars of Creative Taxonomy))
    1. Production
      Macro Close-up
      Studio Lighting
      Selfie Angle
      Frame Pacing
    2. Environment
      Bathroom Vanity
      Gym Workout
      Luxury Interior
      Outdoor Street
    3. Talent
      Gender & Age
      FACS Facial Action Units
      Eye Gaze Focus
      Handheld Velocity
    4. Narrative
      3s Shock Hook
      Problem-Solution
      ASMR Unboxing
      Before-After
    5. Auditory
      Demucs Vocals
      Whisper STT Cadence
      Music BPM Bed
      SFX Audio Ducking
    6. Graphic
      OCR Headline Text
      Sticker Badges
      Subtitle Formatting
      Discount Overlays
    7. Vertical
      Beauty & Skincare
      Fashion & Apparel
      D2C E-Commerce
      SaaS & Apps
    8. Sentiment
      Enthusiastic
      Authoritative
      Urgent
      Calming

Pillar	Focus Area	Example Attributes
1. Production	Camera, framing, and visual aesthetics	Macro Close-up, Wide Landscape, Studio Keylight, Daylight, Cut Frequency (under 1.8s)
2. Environment	Spatial setting and contextual backdrop	Clean Bathroom Vanity, Modern Kitchen, Gym/Fitness, Car Interior, High-End Studio Void
3. Talent	Creator demographics, facial expressions, and motion	Female 25–34, FACS Action Units (Joy, Surprise, Skepticism), Direct Mirror Eye Gaze, Hand Glide Speed
4. Narrative	Structural script format and hook archetype	3s Shock Hook, ASMR Sensory Swatch, Before/After Split, Listicle 3-Reasons-Why, Founder Story
5. Auditory	Spoken cadence, vocals, music, and sound design	Whisper STT Cadence (160 WPM), Demucs Isolated Vocals, 124 BPM Upbeat Pop, -12dB Audio Ducking
6. Graphic	On-screen text, overlays, and typography	OCR Headline Coordinates, 5-Star Rating Badges, Bottom-Third Subtitles, Discount Promos
7. Vertical	Product category and commercial positioning	Clean Beauty, Luxury Skincare, High-Pigment Cosmetics, Activewear, Mobile App SaaS
8. Sentiment	Emotional tone and vocal inflections	High-Energy Enthusiastic, Calm ASMR Whispering, Urgent Scarcity, Trustworthy Clinical

3. Analysis Detail Studio Tabs

The /analysis/[id] detail view provides 5 granular diagnostic inspection panels:

1. Overview Tab

Executive Creative DNA Bar: Top summary badges displaying Primary Hook, Environment, Talent, and Audio score.
TreeSHAP Marginal Lift Indicators: Real-time statistical contribution score (e.g. +27.6% Hold Rate Lift).
Entity Detection Summary: Consolidated breakdown of identified products, logos, faces, and text overlays.

2. Visual & Motion Tab

60 FPS Hardware-Accelerated Playhead: Frame-accurate bounding box overlays with active yaw rotation, velocity vectors, and confidence telemetry.
Forensic Spotlight Defocus: Smooth background blur isolating the exact bounding box subject with zero pixel clipping.

3. Audio Insights Tab (Demucs & Whisper)

Demucs Track Isolation: Independent waveform visualizers for Isolated Vocals / Speech vs. Background Music Bed.
Audio Ducking Telemetry: Detects when background music volume drops to emphasize spoken value propositions.

4. Text & OCR Tab

Interactive Transcript: Spoken words timestamped down to the millisecond.
Soft-Delete & Ignore Action: Fades low-confidence OCR watermarks or accidental text detections, instantly removing them from downstream tag clouds and heatmaps.

5. System Telemetry Tab

Compute Meter: Itemized compute token cost, pipeline runtime duration, and GPU model versioning.

4. Video Timeline DNA vs. 4-Zone Spatial Still Ads

AdHelix evaluates both motion video and static image assets with specialized interfaces:

Video Creatives (Timeline DNA): Temporal multi-track stack (Scene Track, Speech Wave, Music Track, Multi-Tags) synchronized with Meta retention decay curves.
Static & Image Ads (Spatial Layout View): 4-zone spatial grid evaluating Headline Zone, Visual Hero Anchor, Social Proof / Reviews, and Offer / CTA Button.

Creative Analysis & 8 Pillars

On this page