This truly is a hands on tutorial. I do have a question. If you want to interact with the scene in blender, like something flying through the cliff. Why not just take a hold-out shader, making everything behind the mesh the original video again. This wouldn't require any textured output in Blender, and it will just mask the desired occlusion.
If your camera is not in Meshroom (Alicevision) database, and some parameters are missing in EXIF data of picture - you will get nothing on output of Meshroom. I used my phone (Xiaomi Poco X3 pro) and are struggling to get any sane result. Project is very small, just object on the white table, and Meshroom is wrong with all camera positions.
My original model was ~500 mb, the decimated model is ~8mb, but the textured decimated model is ~450 mb using "basic" unwrapping. The other two unwrap methods just produce black textures. Should I expect the textured decimated model to be so big if I'm still using basic unwrap?
So basically you use mesh room not for the 3d scan really, but for getting a tracked camera matching? The one included in blender is very time consuming. Is this a better method?
In this instance, i actually used it for both. I ended up using the mesh as a shadow catcher for a missile i animated into the scene later on. The full video is here - ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-DNTXFEVPwgs.html But yeah, you can use it for a simple camera solve, however, i prefer to do my camera solve in after effects, and import the data using the "AE2BLEND" add-on which can you see in this video - ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-wPb4umDyDDA.html
Unfortunately it won’t work on moving objects, or at least none that I’ve tried! Your best option would be to create a separate model of your object and animate it back in. For instance, if you have a car and a road in your shot, shoot the video clip of the road, then take a separate video/stills of the car, create a separate model of each an composite it back into the scene/footage. If you can’t go back and shoot it, then remove the moving object altogether use CAF in AE (or projection mapping works too) and find a similar model online of your object. If you got an HDRI, matching the lighting should be much easier 🔥
Try aligning your view to a certain angle. Press 7 on your number pad and it will snap your view to the top view, then click and hold your scroll wheel and move your mouse to snap out of the view. If I ever mess up my viewport for any reason, I do this and it quickly solves it :)
The 360 image isn’t a true HDRI, as it’s only taken with one exposure. There are apps out there that claim they can but I’ve never had great results with them. If you want a true HDRI, that’s also seamless then you would be better off investing in a 360 camera like the insta 360. A 360 photo off your phone will still give the same results in terms of reflections though if that’s all you needed, and if enough people ask then I’d be happy to break down my process. Although there’s really not much to it!
Is there a light and non-professional software to align the image if the projector is placed sideways with the hdmi cable? (for reasons of space I cannot project with the projector straight ...) I have a mini-projector the dragontouch mini z1 which with android 7.1 allows me to position the projector where I want and to have a large or small square or rectangle, (the problem is that when I connect the hdmi cable, the image resets itself, deforming ... obviously). A software for the hdmi connection or that makes the image I set the size of the device appear ??? (it has keystone correction but it only works if everything is centered, plus the keystonec does not allow me to shrink the image and change the shape)
