Hi, I have been using Lumafusion to manually stabilize video footage. This works amazingly well, given enough patience and some reference points in the footage. With some small changes to Lumafusion the process can be much improved. The most serious current “flaw” I found has to do with tilting, hence the topic title.
Doing manual stabilisation works even if you introduce some “black borders”, which automatic stabilizers tend to avoid (often seriously shrinking the usable video frame size). Introducing black (out of frame) borders is a compromise, but sometimes the only way to include some non-stabilized footage.
Four suggestions that would really be of extra help:
1. If certain stabilized parts contain “out of frame” (“black border”) segments these segments are see-through. This means if you stabilize footage during a gradual transition from one video source to another you will have a hard time covering-up these borders in an inconspicious way. It would be good if you could choose a color (usually black) for such moments, in stead of see-through.
2. It would help a-lot (!) if a user could place some temporary reference dots on the video display area to be used as a reference. I currently used reference points in the footage that were present near the border/corners, which I was fortunate to have. In its simplest form these reference dots need not be too much aware of context, as long as they can be used for proper framing/tilting etc.
3. Tilting currently seems to work in 1 degree steps. Fine for most circumstances perhaps, although tilting by 1 degree can be very much visible. I had the hardest time covering up such steps at moments where I had to do many edits with panning to keep the frame steady. Could these tilting steps be allowed much more gradual? Ideally you would want tilting to be separate from panning in the case of video stabilisation, but already the option to not trunk tilting values would help a great deal. I believe if you introduce extra (panning) edits during a gradual sequence of tilting each will trunk the current tilting value. If so, please don’t do that anymore.
4. Precise tilting control already is a cumbersome process in its current form. In most photo-editors it is done by pressing your finger anywhere outside of the footage, and move your finger up or down. This is far more precise than moving your finger in a tiny dedicated space.
Several of these suggestions would be of benefit to other usage.