3D模型与视频特效

Run the buddy-cloud.py script with the 3d command followed by your text prompt, for example a Chinese description up to 1024 characters. The script submits the job to the Tencent Hunyuan API and polls until the GLB and OBJ files are ready, typically within 1 to 5 minutes.

Use the video-fx command with a template name and one or more image URLs, such as --template hug --image URL1 --image URL2 for two-person interactions. Over 60 templates are available covering transformations, physics effects, and style conversions.

No, image generation and text-to-video tasks are explicitly out of scope. Image generation must use the ImageGen tool and video generation must use the VideoGen tool; this skill only handles 3D models and template-based photo animation effects.

By default the service returns GLB and OBJ files, and you can request additional formats like STL, USDZ, or FBX via the result-format option. Options also include PBR materials, LowPoly mode, white-model geometry, and adjustable face counts from 10000 to 1500000.

The script has built-in polling that checks job status every 5 seconds for up to 600 seconds, so long waits are normal. If your execution tool times out, resubmit with --no-poll to get a job ID, then query it with the status command instead of resubmitting the task.

The workflow generates a viewer.html using the model-viewer web component and serves it through a local HTTP server on port 18899. Opening http://localhost:18899 in a browser lets you rotate and zoom the model interactively, since file:// URLs cannot load GLB files due to browser security policies.