DVP Tools / Visual Intelligence

Decode Cinematography with FrameDNA

FrameDNA is an optional Prepare tool in Prepare → Image → Video → Vault. Decode a film still into a visual fingerprint, then carry that look into a reviewed image and video workflow.

Start scanning Quick start DNA Methodology

Quick start

Scan your first frame in three moves

FrameDNA bridges the gap between seeing a "look" and describing it to an AI model. It provides the technical vocabulary needed for high-end generation.

Upload the Reference

Drag a film still into the scan zone. Use a high-quality frame that carries the exact Color Grade and Lighting Style you want to replicate.

Extract the Fingerprint

Select your target platforms and hit Scan. FrameDNA identifies dominant colors, luma zones, lens choices, and atmospheric textures in seconds.

Copy the Translation

Review the DNA Grid for accuracy, then copy the platform-optimized prompts. These are calibrated specifically for Kling, Runway, or Midjourney's unique syntax.

Methodology

Visual Deconstruction

FrameDNA reads your frame two ways. Claude describes what it sees, and your browser measures the pixels. The two sit side by side, so you can check one against the other.

The Five DNA Strands

Every scan deconstructs the frame into modular data cells, ensuring total coverage of the visual intent.

Palette Dominant hex codes and color temperature.
Lighting Direction, intensity, and shadow character.
Optics Simulated focal length and depth of field.
Texture Film grain, bloom, and atmospheric haze.
Mood Emotional weight and narrative tone.

Confidence

Claude can give each reading a confidence score. When it doesn't, FrameDNA shows a dash rather than making a number up. A low score usually means the lighting is mixed or subtle, so check that line against the measured panel before you rely on it.

Tip

You need your own Anthropic API key. It stays in your browser, and the frame goes straight from your browser to Claude. The measurements cost nothing and never leave your machine.

Controls

Mastering the Scan

Single Frame Mode

The standard workflow for extracting style from a single hero image. Best for building the foundation of a new project's visual bible.

Compare Mode

Upload a Reference and a Comparison frame. FrameDNA performs a delta analysis, identifying exactly where your footage differs from the target look.

Scene Mode

Drop up to six stills from the same scene, or use a clip and FrameDNA pulls six frames from it on your machine. Name the look, hit ◆ Read the Look, and you get one look for the whole scene: what Claude sees, what the pixels measure across every frame, and what changes from shot to shot. It is one Claude call for the lot.

Platform Adapters

Pick Kling 4.0, Runway Gen-4.5, Veo 3.1, Midjourney V8.2, Grok Imagine or Wan 3.0. The prompt is rewritten in the shape that model expects, with its own aspect ratios.

Named Looks

Save a single-frame scan under a name in this browser and load it back later. For a look you want in other tools, use Scene mode and save it to your Vault.

Measured From the Frame

Next to Claude's reading, FrameDNA measures the frame itself: black and white levels, contrast, saturation, warm or cool, the colour in the shadows and highlights, letterbox bars, vignette and the main colours. Bars are left out of the numbers, so a letterboxed still doesn't read as crushed.

Save to Vault

In Scene mode, ★ Save to Vault puts the look in your Vault: the reading, the measurements and what to carry over. The stills stay on your machine. Save the same look again and Vault updates it rather than making a copy.

EXIF Extraction

If your source file contains metadata, FrameDNA extracts real-world ISO, Lens, and Aperture data to ground the AI generation in physical reality.

Preview This Prompt

Render the extracted prompt before you commit a real generation. Choose Still · Flux or Motion · Kling, hit Generate Preview, then Re-roll until the look lands. Roughly $0.03 a still, billed to your own fal.ai key, so add one first.

Production Packet

Create Production Packet turns a scan into a structured handoff: platform, preset, positive and negative prompts, lens, lighting, palette, mood, action, continuity and delivery notes, with lineage recording where it came from.

Send Packet to CinePrompt

Push the packet straight into CinePrompt instead of copying fields by hand. The camera, lens and grade arrive already populated, so you start from the reference frame's real settings.

Export JSON

⤓ Export JSON writes the whole scan out as machine-readable data, for your own pipeline or to archive a look alongside the project files.

Professional Workflow

The DNA Pipeline

The Reference Bridge

Never start with a blank prompt. Scan a frame from a film you love, then use the DNA results as the 'Style' block for your own scene descriptions.

Grading Alignment

Use Compare mode to see if your AI-generated clips match your initial reference. The 'Delta Report' tells you exactly which prompt variables to tweak.

Look-Book Creation

Read a whole scene from a reference film in Scene mode. ⤓ Export Look (JSON) gives your lighting and design teams the same notes and numbers you are working from.

One look across a scene

Save the look to your Vault once. In AI Studio, Match a film look… writes it into a scene's look fields, so every shot in that scene starts from the same grade, lighting and grain. It guides the model; it can't guarantee every take lands.

Hand off without over-claiming storage

Send the production packet to CinePrompt during Prepare, approve an image before Video in CastLoop, then save the prompt recipe to Vault. The current Vault handoff does not store the generated media attachment.