Kru Batch
over 1 year ago
Hey,
Sorry to hear you're having issues. Two things to keep in mind with this add-on, the first is that it works best with the GPT-4 (if you have access) or GPT-3.5 models. The second and more important point is that the effectiveness is strongly dependant on the user-given prompt.
If the prompt contains detailed instructions of a sequence of actions to execute within the context of Blender, the model should have very little issue producing the necessary instructions internally. If the prompts are potentially vauge or perhaps using terminology not commonly used in the Blender systems, it may struggle more.
Keep in mind that this add-on is for automating tasks within Blender, it's not intended for content generation; if you tell it to design an entire scene, it's not going to do that (or at least not very well).
Bad prompt: "Design an underwater temple scene with caustic light effects, seaweed foliage, and volumetrics"
Good prompt: "Add a cube and a camera to the scene. Add a tracking modifier so the camera is always facing the cube. Add a material to the cube with a pink emissive and enable bloom on the scene, and then render."