Hello!
I know almost nothing about graphics and animations, I am machine learning engineer. I trained a model that predicts animation from voice. Now I have a hard time trying to put my predicted animations (.csv file) into Sequencer so I could add audio track, check that it looks synchronized and render a video file.
To make train-test data, I capture animations with Live Link and MetaHuman model and record animations in Animation Blueprint editor for MetaHuman Face node - “Face_AnimBP”. I also get audio from the video that Live Link Face app saves.
As a result of the recording in Unreal, I get Animation Sequence which I can trim (remove pauses in beginning and ending of animation) and use in Sequencer with any MetaHuman model. I also synchronize audio with animation in Sequencer: I just move audio track until it looks synchronized and then trim audio correspondingly.
Then I export Animation Sequence as .fbx file, and parse it with Python script. I get .csv file with values of animation curves as a result. CSV file looks like:
| No | CTRL_expressions_noseWrinkleL | CTRL_expressions_noseWrinkleR |
|----|-------------------------------|-------------------------------|
| 0 | 0.085 | 0.063 |
| 1 | 0.086 | 0.062 |
I get a .csv file with the same format as a prediction from my model. I have an Animation Blueprint that reads the .csv file and modifies curves of the model according to the file. It works, I can see animations when open it in the editor.
Is there a way to get an Animation Sequence, not Animation Blueprint?
My Event Graph:
My Anim Graph:
