https://x.com/i/article/2116118908860264448

【X 长文】Build a Cyberpunk Scene with Tripo Agentic Canvas in 20 Minutes

Overview

Agentic Canvas is a node-based Agent in Tripo Studio that helps you turn an idea, reference image, or existing asset into a visible, controllable, reusable 3D workflow. You can inspect intermediate results, control how inputs relate, make local changes, and reuse completed work. Starting from one reference image, this tutorial shows how to initiate asset generation with natural language, then use AI 3D Lab to assemble the scene, adjust the lighting, and output an Unreal Engine 5 (UE5) project.

- A Studio-grade creative partner: Simply chat with the Agent to bring your ideas to life as production-ready 3D assets.

- Scalable Batch Workflows on Canvas: One workflow, multiple assets. Batch processing and reusable pipelines, without repetitive setup or switching tools.

- Extend with AI 3D Lab: More creative possibilities in one canvas, from asset inspection and refinement to rigging, editing, and experimentation.

Before you begin

To follow along, prepare a Tripo account and a scene reference image. You can use the image below, which I generated some time ago, or create your own scene concept image in Tripo. The image-generation prompt used in this example follows. If you want to complete the final project import step, you will also need Unreal Engine 5.

Prompt:

Step 1: Create a canvas and upload the reference image

Go to studio.tripo3d.ai. Select Agentic Canvas from the top menu, or use the entry below the image upload panel.

Click Create New Canvas to enter a blank canvas. The following input, generation, and processing steps will appear here as nodes. You can edit nodes manually or use the Agent chat on the right to arrange tasks.

Add the reference image using any one of these methods:

- Click the + icon in the Agent chat input and select the image file.

- Double-click an empty area of the Canvas and choose Upload from the floating menu.

- Drag the image onto the canvas, or paste it directly into the canvas with Ctrl+V.

Step 2: Describe the goal and let the Agent build the workflow

Attach the reference image in the Agent chat on the right, then paste this request. It specifies the scene, which model to use for each type of asset, and the desired output:

This request separates the main architecture from smaller props. H3.1 is used for the main scene to retain richer geometry and surface detail. P2.0 (Smart Mesh) is used for props such as signs and trash cans, where polygon count is easier to control. The Agent plans the steps and connects the nodes; you can inspect each generated asset on the canvas.

Before you submit, choose an Agent mode. Control pauses at key steps so you can review and adjust the results. Auto continues with fewer confirmation points. This example uses Auto. If this is your first attempt, start with Control and continue after each important result looks right.

Step 3: Assemble the scene in AI 3D Lab and prepare the export

Once the models are generated, AI 3D Lab uses the reference image to assemble the assets and adjust their layout and lighting. In this example, AI 3D Lab then exports project files and an import script for UE5. Review the preview in AI 3D Lab for building placement, prop scale, and lighting. If anything looks off, describe the specific changes in chat before generating the output files.

“To continue editing in another digital content creation (DCC) application, open the Export panel in the output area, click Send to, and choose an application such as Blender or Maya. Follow the connection and import steps for that application.

Advanced refinement: If you want to improve the scene further before exporting, send the following request to AI 3D Lab. Focus your feedback on composition, model placement, lighting, and materials so you can assess each revision.

Step 4: Open and inspect the scene in UE5

Download and extract the project archive from AI 3D Lab. Follow the included instructions to open the UE5 project and run the bundled Python import script in the specified environment. In this example, the script imports the models and lights. Check asset placement, materials, and lighting, then edit or render the scene as needed.

Summary

Tripo Agentic Canvas is especially suitable for people new to 3D creation. Once you have a reference image, describe the scene you want in one sentence, and the Agent can help plan the steps, connect nodes, and move generation forward. You can review the results and request changes as you go, starting the creative process with a clear idea.

This tutorial shows an example of starting creation with one sentence and building a cyberpunk scene in about 20 minutes. You can continue refining specific parts through chat and import the result into UE5 for further editing. The node-based canvas preserves the connections between steps, so you can reuse the existing workflow for a similar scene next time and reduce repeated setup.

Open studio.tripo3d.ai, bring a reference image you like, and write your first scene description to start building your own 3D world.