Hello, I am trying to create a fly through video of a recent project and am experiencing playback issues. The video seems to be lagging in that the model is not entirely visible at time of playback. The tiles appear / disappear as the video progresses. I have successfully created fly through videos before with no issues. I even re-processed an old job, created the video again and did not have this issue. What can cause this? I have tried going through all the settings but cannot seem to correct the problem. Thank you for your help!
Could you save your project, close RealityCapture and reset it? Do to do, hold SHIFT on your keyboard and run RealityCapture. You should see a Reset window. Choose Make it like a clean install and click on Reset. After this you will be prompted to sign in again.
It seems there is an error regarding missing parts in the video, but it still hasn’t been fixed. Solutions (or workarounds) were to either create a non-georeferenced model, or to scale it down so it matches a real-life size more. It seems that making a non-georeferenced model works in the most cases.
To create a non-georeferenced model, you will have to run alignment again and create a new model, but this time with disabled prior positions. To do so, select on image in 1Ds view, and press CTRL+A. Now all images should be selected, and you should see Selected inputs panel. Expand Prior pose and change Absolute pose to Unknown. This way geo-location information won’t be taken from images, and now you can repeat process.
Thank you, Ondrej. I will give these a try. Just so I understand correctly, is this a bug within Reality Capture? I guess I’m confused as to why it is not working with this model but worked well with no issues with another georeferenced model on a previous project.
I’m actually experiencing the same issue–parts not loading during the video rendering. I followed your advice to do the “Reset” and now I’m also getting the Capturing Reality logo in the upper right corner, even though all of my images are licensed in the PPI version.
I’m trying to avoid testing the new non-georeferenced alignment & mesh (12,000+ images) since georeferencing is kind of a key component of this project. Having to align & mesh & texture 2x would drastically upend the workflow.
Any new thoughts on this bug and alternative remedies?
I was using the identical settings to what VentureUAV (op) was using. I also licensed via PPI all of the images. The only thing I could think of otherwise is that because I’m pushing a lot on my RTX laptop, maybe I’m just hitting the limits of the program. I did notice that when I simplified the “flight path” of the camera, I was able to see a little more.
Hi Rob, not sure what could be there. As I mentioned this is already written as a bug. So our developers are looking on it. Sorry for the inconveniences.
Sometimes it is a good idea to export the model and create video in another software as RC is mostly for creating models and orthos. But we will look at this issue.
Dear all, I have the same issue when I am trying to export my video. The tiles are not loading properly or not showing at all. When I use another project with the same setting it works. Any idea whats wrong? Thanks!
Hi Ondrej,
I am currently working with a project that has 16500 pictures. All images are geotagged with rtk. it would not work without georeferencing the scene. It would be amazing, if there is another solution. I would love to render the video in RC, because the model is huge and not suitable for importing it to blender. Thanks you so much.
