I was looking to create a more productive pipeline for generating many simple 3D assets
without needing to supervise every one in the web UI. My own use case is mostly for things
like props, background objects, prototype items, or placeholder assets stuff where speed is
more important than getting final quality.
What I am sketching right now is something like: Local store for prompts/reference images;
batch job submissions; store task IDs; poll for progress; retry failures; process exports before
sending to Blender, Unity or Unreal. Overall, it is just the goal to not do everything by hand
one asset at a time.
This led me to explore the API of Tripo AI, which has support for both text-to-3D as well as
image-to-3D very suited for something like this queue-based workflow. I peeked into a few
others as well, but most felt more suited to one-off generations in the UI than batching
submission within a pipeline.
The part that I am less certain of still, is whether those time savings stand up once cleanup
gets factored in. Batch generation sounds great on paper but the gain starts shrinking pretty
fast if too many outputs then need scale fixes and texture correction or mesh cleanup.
Did any of you here in the room build something similar? I wonder what the real bottlenecks
became for you generation time, failed jobs, or just the amount of cleanup afterwards.
[link] [comments]