Raelume template · World BuildingWorlds (Gaussian Splatting) to Scenes
An example workflow on how to use world blocks (Gaussian Splatting) and convert snaps from the world into scenes
Nano Banana Pro
Marble 1.1
Flux 2 Max
Hunyuan3D v3
Seedance 2 Reference
Z-Image TurboNano Banana 2Explore the canvas. Every block, already connected.
What this workflow does.
A concrete workflow example that converts world blocks using Gaussian Splatting into a sequence of scenes. Start from a world block, generate prompts and images, assemble frames, and produce a 3D scene and a video scene sequence. The flow maps inputs (world blocks, prompts, frames) to outputs (new images, a 3D block, and a scene video), illustrating a production AI pipeline for scene creation.
Step by step. The way the blocks run.
- Nano Banana ProGoogle
- Marble 1.1Marble
- Flux 2 MaxFlux
- Hunyuan3D v3Hunyuan3D
- Seedance 2 ReferenceByteDance
- Z-Image TurboAlibaba
- Nano Banana 2Google
- 1Define the world and initial promptsStart with a world block (Gaussian Splatting) and an initial prompt to set the scene concept.
- 2Iterate prompts and generate framesCreate new prompt blocks and generate Frame 1–4 images to explore the scene from multiple angles.
- 3Create 3D asset from imagesConvert the best frames or combined inputs into a 3D car or other assets using a 3D block.
- 4Assemble into a scene sequenceBring together images, 3D assets, and a video sequence to form SCENE 1 and related clips.
- 5Produce final scene videoAdd a New Video Block and a video-sequencer to render the final scene as an output video.
Common questions. Straight answers.
Outputs include New Image frames, a WORLD block, a New 3D Block, and SCENE 1 video sequence.
Inputs include the Desert/world block, prompts, frame images, and 3D assets used to build the final scene.
Yes. You can adjust prompts and regenerates frames (Frame 1–4) to refine the scene before finalizing the video.
3D blocks derived from image inputs are incorporated into the scene to enrich the final render and perspective.
The workflow uses a mix of image and video blocks, 3D blocks, and world blocks from compatible AI tools to produce the final scene.
