Ray 3.2: Overview
Ray 3.2 is a production grade Video to Video (V2V) transformation tool designed to restyle existing footage while preserving its core structure. It starts with a source video and outputs a re imagined version matching the exact same duration.
Best Uses:
Restyling live action into anime, claymation, or painterly aesthetics.
Updating product colors, materials, branding, or finishes.
Swapping environments, skies, weather, and wardrobe.
Creating regional, seasonal, or demographic campaign variants.
Polishing pitch decks, temp VFX comps, or concept art directions.
Key Differences: Ray 3.2 vs. Ray 3.14 Modify
Feature | Ray 3.14 Modify | Ray 3.2 Modify |
Keyframe Support | Start and end frames only | Up to 64 keyframes at custom timestamps |
Duration Handling | Trims or loops to 5s / 10s outputs | Matches source duration exactly (up to 20s) |
Production Exports | Standard output | Supports HDR video and EXR export for color grading |
Processing Modes | Single mode | Choice between Speed vs. Quality modes |
Rule of Thumb: Use Ray 3.14 only when you specifically need to loop or change clip length. Use Ray 3.2 for precision production work, deep character controls, multi keyframe alignment, and professional color workflows.
How to Access Ray 3.2
Right Click (Recommended): Right click any video asset > Modify Video > Ray 3.2 to launch the complete modal interface.
Agent Chat: Select your video and keyframes on the canvas, then instruct the Agent in plain English.
Toolbar: Go to Generate Video > Ray 3.2 to open the simple generation modal.
The Three Core Inputs
Ray 3.2 requires a Source Video and at least one of the other two inputs to run.
Source Video (Required): The baseline footage up to 20 seconds long. Output duration always matches source duration.
Prompt (Optional with Keyframes): Describes the target final look. Prompt for the desired end state, not the transition process over time.
Keyframes (Optional with Prompt): Up to 64 custom guide images placed at precise source timestamps to anchor specific moments in the timeline.
Adherence Controls (Structure & Motion)
Adherence settings dictate how strongly the output protects the source material. Both settings use a numeric scale from 1 to 9 (9 being the strongest preservation level).
Motion (1 to 9): Controls movement fidelity. Higher values keep timing and movement paths closer to the original. Lower values allow the model to reframe how movement feels.
Structure (1 to 9): Controls spatial arrangement. Higher values protect camera framing, scene layout, and object geometry. Lower values let the model reshape the scene layout.
Character Controls (Faces, Bodies & Poses)
Character controls act as independent locks to preserve identity and movement through transformations.
1. Faces
When ON: Locks facial features, expressions, and geometry. Use for dialogue, emotional close ups, or when a specific actor must remain recognizable across style changes.
When OFF: Use when swapping a character into a robot or creature, or when faces are distant or off screen.
2. Bodies
When ON: Preserves the subject's overall silhouette, proportions, posture, and build. Ideal for wardrobe changes, sports, or dance clips.
When OFF: Turn off when altering body proportions entirely (such as changing a human into a smaller creature or bulky robot).
3. Skeletal Tracking Modes (Poses vs. Blocking)
Poses: The heavy, highly adherent tracking mode. Use for detailed physical performances (such as intricate hand movements or complex choreography).
Blocking: A lighter, sparser tracking mode that follows key joint positions without connecting lines. Best for looser motion, abstract characters, or non humanoid creatures.