Would be interested to see VEAI compute and add motion blur into footage, particularly CG footage.
Something akin to the ‘ReelSmart Motion Blur’ plugin but using machine learning instead of optical flow.
I am curious: What would be the benefit of an AI based motion blur instead of a conventional one? Could you elaborate ?
The optical flow technique falls apart when applied to fast moving footage. I’m curious if an AI based approach would reduce the amount of glitches.
Thanx for getting back. I´ll have to twist my head around this - the AI would have to take the time domain into account, not sure at the moment if a neural net approach would outperform something with motion vector detection - but it sure is an interesting thought…
Maybe i´ll dig through the papers of the recent status of motion blur synthesis and removal (which also is interesting)…
This would actually be very interesting to use on old TV show footage from the 90’s like Star Trek Voyager. The CGI footage is quite jittery and has no shutter angle so the video is choppy. I’ve seen the original series remastered on Netflix and it has motion blur that looks extremely realistic but I have a feeling that the algorithms used to do that are not even close to free and probably would be easily outperformed by an AI model.
The current Chronos AI model already produces higher quality intermediate frames than anything Adobe can do. Currently the best way to add artificial motion blur is with After Effects with the Pixel Motion Blur effect. I think it would be great to see what VEAI could do on this front. High shutter speed 24/25 fps smartphone video would look much more professional.
Edit: Technically you can already do this with some extra work. For example if you would interpolate a clip with the Chronos model at 2000% you would have 19 extra sample frames. If you import and interpret that clip in a program like Premiere Pro at its intended framerate and then speed it up by 2000% with Frame Blending you could get a similar result.
I totally agree, there are some tools on the marked but often they lack from distinguishing static background from moving objects and the object as well as the background is blurred (e.g. RSMB). Chronos does a great job in separating differently moving object layers. Hence, the AI approach should work perfectly here.
I already posted it in another thread, I would love to have a generalized “motion blur control” with a shutter angle control. Topaz guys are already working on a motion blur reduction model, so a full motion blur control (reduce&increase) would be a wonderful completion for a VEAI model.
I would like to see a feature that adds realistic motion blur to footage recorded at a high shutter speed. I strongly dislike having to increase the shutter speed on my camera to compensate for overexposure. However, it is the most convenient way to shoot with a wide aperture without having to purchase expensive ND filters for each lens. The downside of this approach, though, is that it eliminates natural motion blur, which is what I would like to preserve. Hopefully, someday this feature can make its way to the app. 
I think it would be really useful . So far rsmb is considered the best way but it has limitations and doesnt look realistic to me.
An interesting idea.
Blender, when it renders computer generated movies, by default, it has no motion blur. To add it, it has to generate a bunch of in-between frames then blend them. It has absolute control over motion and speed, so it has the information to do it.
For TVAI to do something similar, you could run it though 16X slomo. No idea what program to use to blend the images though. (Blender probably can, but I’m not sure how to tell it to do that. [I would also not be surprised if ffmpeg can do it. In that case, Topaz could make commands to automate it.])
EDIT: I missed the part in the title where you specify high shutter speed. In that case, the in-between frames are probably already there.
This would be great. So much footage shot on drones / iphones etc. are with high shutter speed and are problematic for that reason when incorporating into footage shot with cinema cameras. It would be great to be able to basically ‘remove’ those problems by 1) adding motion blur and 2) remove the excessive sharpening whilst retaining detail.
I already use TVAI to add motion blur to my drone footage by using the mentioned workaround of slowing it down 16x in TVAI and then setting it to 1600% speed in Premiere/After effects with frame blending enabled, often achieving very good looking motion blur and making the footage appear much smoother. It would be great to have this incorporated inside of TVAI directly, with shutter degree control (I believe at a 180 degree shutter setting it only needs to interpolate half the frames, right?) and control over how many interpolated frames are needed (I often do an export at 8x, and then decide if the blur is working or if it needs 16x or even more) thus completely eliminating the need for ND filters on drones and small cameras like gopro’s or smartphones that are often degrading the image as well as flimsy to work with.
