Hello,
I am a research assistant and I help a student with his bachelor’s thesis.
As part of his thesis, he tries to automatically create a MetaHumanIdentity mesh based on footage.
We found the example script create_identity_for_performance.py (located in the plugins Content\Python folder) which already accomplishes some of these steps and tried to scale it up for multiple promoted frames. But as soon as we start to track the second frame (MetaHuman_identity_asset.start_frame_tracking_pipeline(…)) we get the error message The promoted frame processing failed. every frame processes fine after that (the generated indentiy has no tracking data for the first frame, but every other frame has tracking data). We gathered the detailed error message (with MetaHuman_identity_asset.diagnostics_indicates_processing_issue()) and it states that Identity face scale is 0x that of an average MetaHuman, which may indicate an issue with the input data. Please check your footage camera calibrations are correct and that calibration distances are expressed in cm., which doesn’t make any sense because if we switch the order of the processed frames around each time only the first frame gets this error. So, the same data shuffled around resolved in different behavior.
In search of a workaround, we tried:
- Waiting until
MetaHuman_identity_asset. is_frame_tracking_pipeline_processing()becomes false, but the first frame never seems to finish (we waited a couple of minutes and nothing happened, while every other frame finishes in humanly immeasurable time) - Just proceeding after the error message and retracking the first frame after the others are finished. Which works perfectly for a total of two frames but above weird thinks happen. (For example if we processed three frames, the first frame got the tracking data of the third, the second worked perfectly and the third had no tracking data)
- Solving the identity after every added frame (with
face.conform()) which resulted in perfect tracking data and no error messages, but takes ages because of the unnecessary solves.
If needed: The student is using UE 5.3.2 with the newest MetaHuman plugin version (3.0.3) available from the marketplace on an windows pc.
Thanks for reading 
1 post - 1 participant