Learning Timeline
Key Insights
Language Issues in the Output
GLM-5 might generate text or labels in Mandarin even if your prompt is in English. You can provide an additional prompt to change the interface language to English if needed.
Fixing 3D Geometry
Sometimes, the generated 3D objects may have visual errors, such as disconnected roof sections. You will need to perform 'fine-tuning' by providing specific prompts to add detail or repair the structure of the object.
Prompts
Basic Prompt for 3D Scene Generation
Target:
GLM-5
I'm going to upload this image and see if it can create a 3D animated scene from that. Use a single HTML file.
Prompt for Enhancing Interactive Lighting
Target:
GLM-5
Allow me to change the time of day and make sure the lighting actually works.
Step by Step
Steps to Generate an Interactive 3D Scene from a 2D Image
- Open the GLM-5 interface and click the 'Upload' button or the image attachment icon.
- Select a static 2D image (e.g., a photo of a building or landscape) from your device.
- Enter a prompt instructing the AI to convert the image into a 3D scene within a single HTML file.
- Click the 'Generate' button to start the coding process.
- Once the code is generated, check the 3D output. If there are language or lighting issues, provide a follow-up prompt.
- Ask the AI to add a time of day control feature so the lighting can change dynamically.
- Click 'Generate' again to update the HTML file.
- Use the slider provided in the HTML file to change the time (Dawn, Noon, Dusk, Night) and use your mouse to navigate and view the scene from various angles.
