What our run establishes
Our run on September 28, 2026 tested commit dee5dae. The matching stored sandbox record identifies a Python 3.12 Bookworm image with 3 CPUs and 8 GB RAM. Installation took 19 seconds; the build step took 4 seconds. Both returned exit code 0 without a recorded timeout. These are separate step timings, not a combined installation time.
Our harness records each command outcome through success status, exit code, elapsed time, timeout status and a log tail. Here, the build tail is empty, so the record supports the outcome without identifying what that step exercised. Tests were skipped with the reason "no script". A skipped test step is not a failed suite, and this record supplies no passing-test count.
The attached counts are stored parser summaries, not independently reconciled, exhaustive counts. No new execution or complete-log reconciliation was performed for this correction. The testing methodology explains the publisher's recording approach and limits; its current text does not establish which procedure version the September run used.
What you must connect yourself
Camera to Blender requires a Tripo3D API key for model generation; Gemini background removal is optional. The README describes a camera page, a relay server and a Blender add-on. You run the server on the computer where Blender is open, then use a phone browser or laptop webcam to capture the image.
For phone use, the documented setup requires an HTTPS ngrok address. Blender uses the corresponding WebSocket address. Free ngrok URLs change after a restart, so the README tells users to reconnect Blender with the new address. Laptop webcam use can skip ngrok and connect locally, but generation still requires Tripo.
Start with that documented local webcam route if you want to check the relay and Blender connection before adding the phone tunnel. A successful connection is only an intermediate check: you still need to verify generation and import with your own credentials. Our run did not test those outcomes, and we did not measure generation speed or model quality.
Read the audit as an advisory record
Our sandbox's stored pip-audit summary records 30 advisories across 3 packages. The record does not include scannedPackages, so scan coverage is unknown. Those figures do not establish exploitability, a complete security assessment or the safety of a deployment.
The detector also recorded no CI workflows, Dockerfile or tests directory. Together with the skipped test step, this leaves little automated evidence for the complete workflow. Review the dependency findings and test the actual capture-to-import path before relying on it for repeated work.
The merged provider fix has a testing limit
Pull request #2 replaced deprecated Tripo model versions. Its contributor reported that the new versions reached the credit check, then stopped because the account lacked credit. The contributor explicitly did not complete a generation with credits available. That report is separate from our sandbox results.
The same contributor reported a planned November 1, 2026 shutdown of Tripo API V2 and said the change retained V2. The supplied evidence does not verify whether that shutdown occurred or establish current service compatibility. Check the provider's current API requirements before investing in setup.
Issue #1 documents an add-on dependency failure caused by Blender using its own Python environment. Its closed status does not prove that our sandbox enabled the add-on or that it works with every Blender version.
Alternatives replace generation, not the whole relay
TripoSR provides local inference that saves a reconstructed 3D model to an output directory. Its README documents installation and manual inference. Consider it when running reconstruction yourself matters more than having this repository's camera and Blender connection. It replaces the generation approach; do not assume it plugs into this relay without integration work.
InstantMesh provides a local Gradio demo and command-line mesh generation. Its README recommends Python, PyTorch and CUDA dependencies and describes model downloads. Consider it when you want to work with reconstruction models and can manage that environment. It is another generation alternative, rather than evidence of a ready-made replacement for the phone-to-Blender workflow. Neither alternative was tested in our run.
How to read the scores
Scores use a 1-to-5 editorial rubric: 1 means substantial barriers or very limited supporting evidence; 2 means important gaps; 3 means workable with manual setup; 4 means well documented or well supported within the assessed scope; 5 requires strong evidence across that scope. They are judgments, not benchmarks.
Setup ease is 3 because installation completed but keys, the add-on and connection setup remain. Documentation is 4 for the README's phone and local-webcam instructions and troubleshooting, not proof of current API compatibility. Community is 2 because the supplied issue and pull request show concrete contributions but do not establish sustained support. Maturity is 2 because our record lacks supplied tests and CI, while the provider change has no completed-generation evidence.

