Workflow
Due to the focal length of camera that is 16mm, use length grid of 16mm to calculate the distortion information with Lengthdistortion.Click Detect which could automatically calculate the length distortion.
Add cameratracker to the undistorted clip.Then select refine feature locations and preview features to show the coordinate points to help us estimate. Add the numbers of points from 150 to 800, for the better result.
Click solve to transport the tracking result to available data.
Select the three selections below and delete invalid points.
Create the scene with unlinked camera.
Use writegeo to render the camera to maya.
Copy lensdistortion, select the output as stmap, and add a shuffle to convert the result to a uv map for exporting to stmap.
Import into stmap.
Import the model with the camera already matched in maya and composite it with the scene.
Give the distortion of the lens to the finished model by stmap.
Import smoke and use timeoffset to adjust the frame rate and grade to adjust the color.
Use the defocus node as blur.
Create a new card, and the previous tracking camera to bring over the smoke to match the camera movement, and then merge composite to the scene, composite the way to choose hypot.
Then copy the stamp to the smoke and give it a lens distortion.
After Roto, export a scene with corrected lens distortion for tracking in mocha, and export it in jpg format.
Open mocha and create a new project.
export track
Go back to the frame where the roto was drawn, and adjust the cornerpin.
Given to the scene to create a mask.
Add keyframes to the merge of mask, in 169 to 0, 168 to 1.
For the background to clear markers, use rotopaint at frame 249.
Back to the cameratracker, select the points on the ground create card, and then use this card to replace the card just connected to the ground mask.Draw a rotopaint again, framehold is 150.
Select the tracking points on both sides of the wall at the cameratracker to create the card.
Adjust the project, smoke composite to the top of the wall of roto.Use grade plus roto to add shadows.
Use the cryptomatte plugin to layer and add noise to the steam engine.
Balckpoint extracts the darkest parts of the steamer
White extracts the brightest spot
Then D turns off the grade
Lift extracts the darkest part of the scene
Gain extracts the brightest part of the scene
Then D turns on grade.
Add noise to the material and defocus.
Final and conclusion
Color mixing problem
Use blackpoint and whitepoint to determine the darkest and brightest values of the material, i.e., the shoulders and toes, respectively.
Gain and lift are used to absorb the scene is the brightest and darkest value of the scene to match the material, uniform light.Then use grade or colorcorrect to color match the scene.
Other Problems
- When removing lens distortion, the exported stmap has no content.
It turns out that the stmap is not selected in lens distortion - After importing the model, many parts were not realistic and were not composited into the scene.After the model was toned, using grade and cryptomatte to match the light and tone of the scene.
- I found that there are actually many details need to be adjusted when roto, manual roto is very time-consuming,the appropriate use of some blur, and track the movement of the wall, the movement data match to roto, and then manually adjusted slightly, saving a lot of time.
- During the production process, I encountered the problem that the project was very stuck, for the roto part, and all the stmap match lens distortion nodes are temporarily turned off, it saved a lot of time.
- The problem occured when exported into maya due to the wrong camera choice.
- Initially, when the camera was imported into maya, it was necessary to export a cube in nuke with a predetermined placement, so that the cube could be imported into maya to adjust the model according to the cube, and then export the model.