Reposition and pose any subject in any shot with complete control over angle and composition. This tutorial shows you how to use AI to extract a 3D model from your image, render it from any camera angle, then composite it onto a clean background plate. Standard editing only gives you crop, scale, and warp. This gives you full spatial control.
The Node Pipeline
Open the Workflows tab in Sequencer. You will build a six-node pipeline: the input image feeds into 3D generation, which feeds into the render node for repositioning. Meanwhile, a parallel branch creates the clean background. Everything merges in the composite node.
The pipeline has two parallel tracks. The top track extracts and repositions the subject. The bottom track removes the subject from the original image to create a clean background. Both tracks converge at the composite node.
Step 1: Extract the 3D Model
Start by adding your source image. Drag an Image node onto the canvas and upload your shot. This feeds into the Generate 3D node.
The Generate 3D node uses Trellis to create a 3D mesh from your image. Connect the image output to the 3D input. Run the node. You will see a preview of the extracted 3D model appear. The AI captures the geometry, texture, and form of your subject.
For best results, use images with clear subjects against relatively simple backgrounds. Complex scenes with overlapping elements are harder for the model to interpret accurately.
Step 2: Reposition with Render 3D
Connect the 3D output to a Render 3D node. This is where the magic happens. You now have full camera controls over your subject.
Camera Controls
Theta
Horizontal rotation (orbit left/right)
Phi
Vertical angle (orbit up/down)
Radius
Distance from subject (zoom in/out)
FOV
Field of view (perspective distortion)
Drag in the 3D preview to rotate the camera interactively. Or enter precise values in the text fields. Low FOV values create a telephoto look with flattened perspective. High FOV creates wide-angle distortion.
Set the aspect ratio to match your final output. The node renders a clean image of your subject from the new angle, ready for compositing.
Step 3: Create the Clean Plate
The repositioned subject needs a background to sit on. You cannot use the original image because the subject is still there. You need a clean plate: the same background with the subject removed.
Add a Mask node and connect it to your original image. Paint over the subject to create a mask. The white areas indicate what should be removed. Be thorough. Any missed pixels will show up as artifacts.
Feed the mask into an Erase node. This removes the masked area and fills it with AI-generated background. The result is your clean plate: the original scene without the subject.
Step 4: Composite the Layers
Now you have two images: the clean background plate and the repositioned 3D render of your subject. The Composite node combines them.
Connect the clean plate to Layer 1 (the background). Connect the 3D render to Layer 2 (the foreground). Expand the transform controls to position, scale, and rotate the subject exactly where you want them in the frame.
The interactive preview shows exactly how the final composition will look. Drag the subject around until the composition feels right. Lock Layer 1 to prevent accidental background movement while you adjust the subject placement.
Step 5 (Optional): Refine with AI
Sometimes the composite edges need cleanup. The 3D render might have slightly different lighting than the background. An optional refinement step can smooth these transitions.
Feed the composite output into an Inpaint node. Mask the edge areas where subject meets background. Use a prompt like "seamless blend, consistent lighting" to have the AI harmonize the transition. This final polish makes the composite look natural.
Getting Professional Results
Source quality determines output quality. High-resolution, well-lit images produce better 3D models. Harsh shadows or motion blur carry through the pipeline.
Match the lighting direction. If your original scene has light coming from the left, position the 3D render so its lighting matches. Mismatched lighting immediately breaks the illusion.
Save your workflow as a template. Once you have the pipeline working for one shot, you can reuse it for similar compositions. Swap the input image and adjust the camera angle. The structure stays the same.
Build Your Composition
Open the Workflows tab and build this pipeline. Six nodes, complete compositional freedom. Position your subjects exactly where they need to be.