fcp.co logo transparent
fcp clapperboard
Welcome, Guest
Username: Password: Remember me
14 Nov 2020
Moving over to a new system, we couldn't take the old avatars - so please upload a new one!
  • Page:
  • 1

TOPIC: Resolve To FCPX Bug With 29.97p Clips?

Resolve To FCPX Bug With 29.97p Clips? 04 Jan 2021 21:35 #111903

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
Hello all!

I am trying to cut in the final color pass of a show I am working finishing up.

The project is 1920x1080 23.98p.

Some of the clips in the project are 29.97p.

When I A/B (compare) the Resolve rendered clips to what was already in the timeline every forth frame from the resolve clip matches the original clip in the timeline. The frames in-between those 4 are off. They are off in that they aren't exact matches, like they are being interpolated by either Resolve or FCPX differently.

Both clips are the same frame rate and frame size as well as interpolation (i.e. progressive).

I took the original clip and the Resolve rendered clip and placed them into a Premiere timeline with the same specs as the FCPX Project. In Premiere everything is interpreted correctly. I.e. every frame matches.

This leads me to think that this is an issue with FCPX.

Anyone experience this issue before?

Happy to clarify as well if I didn't explain the issue well.

Brian
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 04 Jan 2021 22:32 #111905

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
One other observation is when I watch the source timecode of the two clips at the same time, FCPX drops frames at different points for each clip. It's like its interpreting the footage for each one differently even though they have the same frame rate, timecode, frame size.

Codec is different: Original is ProRes 422, Resolve output is ProRes 4444
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 04 Jan 2021 22:34 #111906

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: ...The project is 1920x1080 23.98p...Some of the clips in the project are 29.97p.

When I A/B (compare) the Resolve rendered clips to what was already in the timeline every forth frame from the resolve clip matches the original clip in the timeline. The frames in-between those 4 are off. They are off in that they aren't exact matches, like they are being interpolated by either Resolve or FCPX differently...


You have 29.97 clips in a 23.98 timeline.  That requires some rate conforming algorithm, sometimes seen as a periodic dropped frame. The goal is never get in this situation because there is no perfect solution.

As a workaround in FCP you can try optical flow rate conforming. It can often smooth out the required rate conform, but it requires additional computation when rendering. Optical flow conforming can sometimes introduce morphing artifacts, but it generally works fairly well.

To use optical flow rate conforming, select a clip in the timeline with a non-matching frame rate, then in the FCP video inspector, scroll to bottom and under Frame Sampling select Optical Flow. Then select that clip in the timeline, render with CTRL+R, and examine the playback smoothness. If that works, do all the clips with frame rates that differ from the timeline rate. 

I don't quite understand your exact FCP/Resolve workflow, but it's likely this is the underlying problem.

Gerald Undone discusses mixing frame rates in this video. At 09:18, see especially his discussion of using 30 (ie 29.97) material in a 24 (ie 23.98) timeline.

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 05 Jan 2021 14:01 #111930

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
Thanks for the reply Joema. As always I value your input.

I whole heartedly agree that mixing frame rates like this is not great. Unfortunately this is the nature of this particular documentary. Had we been able to control the footage coming in we would have.

I tried out the optical flow like you suggested but didn't see much change. Good to have the tactic in my back pocket though.

I know I am not in an ideal situation ,and with delivery around the corner, I think we will just press on as is. Not much time to trouble shoot.

The colorist and I are trying a few other things, and we'd both like to figure out why it was seamless in Premiere and not in FCPX. I don't think I have enough data points yet to really put to rest that all this issue is is having mixed frame rates.

Once the show deliveries and we have time hopefully I can give a clearer answer as to whether it's just rate conform or some weirdness between FCPX and Resolve.

Thanks again,

Brian
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 05 Jan 2021 15:26 #111934

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: ...The colorist and I are trying a few other things, and we'd both like to figure out why it was seamless in Premiere and not in FCPX...


When you say "the frames in between those 4 are off", you mean the duration or edit points are off, or each corresponding frame shows a slightly different image between FCP and Resolve? If the latter, that is hard to understand because the only thing recorded are the frames. 

If you compare 29.97 material on a 23.98 timeline in FCP to the original file as played in Quicktime, there will be some motion cadence differences due to rate conforming. But Resolve behaves identically to FCP -- they both drop every 5th original frame, producing a timeline where every 4th frame has a bigger movement jump.

Resolve is more aggressive than FCP at trying to change the timeline characteristics. Even if you manually create a 23.98 timeline, if the first clip added is 29.97, Resolve will prompt you for changing the timeline format. If you just hit go on the dialog, it defaults to changing it to match the clip being added.

Is it possible that in your case Resolve is actually using a 29.97 timeline, not a 23.98 timeline? You can check timeline characteristics in Resolve with SHIFT+9.

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 06 Jan 2021 16:49 #111954

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
Hey Joema,

Sorry I wasn't clear. When I cut the Resolve generated footage on top of the original footage and watch the source timecode of both frame by frame. I get this: It appears that FCPX is deciding to drop frames in different places on the clips to compensate for the 29.97p footage in the 23.98p timeline. The clips are mostly in sync it's just the in-between frames aren't exactly the same. So on interviews it is a little wonky looking.

Here is an example you can watch: app.frame.io/r/24556df0-b199-4e97-9e7e-8d92e2986838

To answer your previous question regarding Resolve I sent that along to the colorist and he said he has it in a 23.98 timeline. I don't use Resolve much so I can't comment on it with any authority.
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 06 Jan 2021 17:56 #111955

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: ...your previous question regarding Resolve I sent that along to the colorist and he said he has it in a 23.98 timeline...


(a) What is the original frame rate which later became the "Resolve-rendered clip"? IOW for the clip in question, what frame rate did you send him? You can verify that by playing the submitted clip in Quicktime Player and doing a CMD+I.

(b) What is the current frame rate of the Resolve-rendered clip? IOW that same clip, but the version returned from the colorist. Could you please play that clip in Quicktime Player and do a CMD+I to verify?

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 06 Jan 2021 19:36 #111958

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
Question A - 29.97
Question B - 29.97
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 06 Jan 2021 20:17 #111960

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
To add to the mystery:

Drop frame clips are coming back from Resolve as non drop. But if I reveal in finder and reimport them into the time line they become Drop Frame again.

The colorist says he's not having any of those problems on his end.

I wonder if my project and/ or Library is corrupt at this point.
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 07 Jan 2021 19:10 #111981

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: Question A - 29.97
Question B - 29.97


I did some testing with Resolve and FCP and I don't see the problem. However I am still confused about the exact processing history of each clip.

If the original 29.97 clip is in a 23.98 FCP timeline, that will discard every 5th frame to conform the rate. Thus when you step through it in FCP frame by frame, each 4th frame will show a bigger jump on objects in motion (as expected). 

If the colorist received the same original 29.97 clip and if he edited it a 23.98 timeline, that should show the same behavior. However if he gave you back a graded 29.97 clip, I don't see how he did that from a 23.98 timeline. You can force Resolve to export 29.97 from a 29.97 clip conformed to a 23.98 timeline but it looks bad.

If instead he used a 29.97 timeline, then there would be no rate conforming in Resolve, and his exported graded clip should have the same motion characteristics as the original clip if imported to FCP.

I did the above test, laid the Resolve 29.97 color-graded clip on top of the original 29.97 clip in FCP, and set opacity of the top layer to 50%. That way I could single step and see any slight deviation between frames from the two clips. They remained perfectly aligned.

Now, if by chance when adding the Resolve clip to the FCP timeline as a connected clip, it is positionally mis-aligned by 1 frame vs the bottom clip, then you can see the "every 4th frame" behavior you mentioned. This is because on low-motion shots each clip is only 1 frame off, so looks mostly aligned. Then each 4th frame is the expected rate conforming jerk, but one clip hits that before the other clip due to the positional mis-alignment. 

I suggest you zoom way in on the clip starting point and make certain the first frame of the bottom and top clips are perfectly aligned. You can confirm this by doing right-click on the top clip>Show Video Animation, then pulling down that clip's opacity to about 50%. Then if you single-step, each frame should stay aligned -- assuming they are both really 29.97 and one clip hasn't had unique rate conforming baked in from different processing.

Please Log in or Create an account to join the conversation.

Last edit: by joema.

Resolve To FCPX Bug With 29.97p Clips? 08 Jan 2021 20:08 #111988

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
Here is what the colorist did (pasting in his message to me):"The round trip workflow ignores timeline frame rate and only takes into account when writing the xml. Timebase in resolve only effects the way the resulting sequence is set up in the host NLE. I could be working in a 2000fps timeline, it doesn’t mean Resolve will render bad clips in the round trip output. Using any of the round trip options when going out of Resolve will always render those clips the same, 1:1 with the source. The only bearing timeline settings or the different round trip workflows has on anything is that XML it writes for the host NLE. The clips themselves are preserved and left alone, 1:1 with the source"

I did the above test, laid the Resolve 29.97 color-graded clip on top of the original 29.97 clip in FCP, and set opacity of the top layer to 50%. That way I could single step and see any slight deviation between frames from the two clips. They remained perfectly aligned.

This is the bugaboo. When I place the colorist's delivered 29.97p clip above the original 29.97p clip in the 23.97p timeline. Doing the opacity trick, I can see the deviation. The original clip's cadence drops even frames. The clip that came in via the Resolve xml is dropping odd frames.

Now it gets weirder... If I reveal the colorist's 29.97p clip and the original 29.97p clip in the finder and then drag them back into the 23.97 timeline and overlay them. They match up. There is no discrepancy. Both drop the even frame on the cadence and using opacity everything matches up.
So it appears that it's either:

A) user error (I'll eat crow if it is)
B ) a bad xml out of Resolve
C) FCPX isn't translating the xml properly.

Those are my current findings.
Owner InnerParakeet
www.innerparakeet.com

Please Log in or Create an account to join the conversation.

Last edit: by bcassin.

Resolve To FCPX Bug With 29.97p Clips? 09 Jan 2021 14:51 #111994

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: ...This is the bugaboo. When I place the colorist's delivered 29.97p clip above the original 29.97p clip in the 23.97p timeline. Doing the opacity trick, I can see the deviation. The original clip's cadence drops even frames. The clip that came in via the Resolve xml is dropping odd frames...


I just did that test, and I don't see any deviation. Procedure: I took a UHD 4k/29.97p clip from a Panasonic S1 which was captured at as 500 mbps ProRes 422 on a Ninja V, imported it to FCP, then imported the same clip to Resolve on a 23.98 timeline, color-graded it, then used the option to export to FCPX. That is different from the normal Resolve export which rate-conforms the clip to the timeline frame rate. The FCPX export option does not rate conform the clip, so your colorist is right about that. My previous statements assumed the regular export method.

The export produces a 29.97p graded clip and XML file for FCP. I imported that to FCP, stacked that and the original clip on a 23.98 timeline, set opacity of top clip to 50%, and single-stepped through the clip. I didn't see any deviation or alignment mis-match. As expected they are both rate-conformed from 29.97 to 23.98 which produces a jerky motion every 4th frame if subjects are in motion, but they are both handled identically.

Using FCP 10.5.1, Resolve Studio 16.2.8 on an iMac Pro running Catalina 10.15.7 and Apple Pro Video Formats 2.20.

At this point you need to examine the exact characteristics of the original clip and the returned Resolve clip. If one is interlaced, was somehow interpreted in Resolve or has some odd characteristic, that might account for it.

The best way to do this is using the data comparison tool Invisor, which allows spreadsheet-like side-by-side comparison of video metadata from multiple files:  https://apps.apple.com/us/app/invisor-media-file-inspector/id442947586?mt=12

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 11 Jan 2021 22:49 #112019

  • bcassin
  • bcassin's Avatar Topic Author
  • Offline
  • Senior Boarder
  • Senior Boarder
  • Posts: 55
  • Thank you received: 0
I used the Invisor app and they appear to be identical. I've attached a screen shot.

The colorist's system is running these specs:
FCPX 10.5.1, Resolve Studio 16.2.7.010, Catalina 10.15.6, Mac Pro (2019) 24-Core / 256GB RAM / AMD Radeon Pro Vega II Duo 32GB, Apple Pro Video 2.20

I'm running these specs:

FCPX 10.5.1, Catalina 10.15.7, iMac (Retina 5K, 27-inch, 2019), Apple Pro Video 2.1.3

Perhaps its that I have an older version of Apple Pro Video. Which I am hesitant to upgrade till after I deliver this show.
Owner InnerParakeet
www.innerparakeet.com
Attachments:

Please Log in or Create an account to join the conversation.

Resolve To FCPX Bug With 29.97p Clips? 12 Jan 2021 13:19 #112028

  • joema
  • joema's Avatar
  • Offline
  • Platinum Boarder
  • Platinum Boarder
  • Posts: 1775
  • Karma: 27
  • Thank you received: 395

bcassin wrote: I used the Invisor app and they appear to be identical. I've attached a screen shot...


Thanks. Since Invisor indicates the clips are actually 59.94 fps, I re-ran the previously-mentioned Resolve>FCP export test using 59.94 ProRes 422 source material, and it behaves the same. IOW I don't see any alignment mismatch vs the original clip after loading the Resolve-generated XML and 59.94 clip back into an FCP 23.98 timeline.

As expected, since 59.94 is not an even multiple of 23.98, FCP must rate conform both original clip and the Resolve clip. But it does this identically for both stacked clips, hence no mis-alignment.

There's a remote possibility a difference exists due to Pro Video Formats 2.20. However I think it's more likely some kind of processing or rate conforming took place in Resolve which produced a different cadence output at the same frame rate. 

There are many different rate conforming algorithms. You could have two different algorithms which conform the same source and destination frame rates, yet the resulting cadence could be slightly different.

For some conversions, a certain conforming algorithm might be more common. E.g, 3:2 pulldown is commonly used for conforming 24p to 29.97i.  I don't know the conversion algorithms used for 59.94p>23.98p, but it's possible different NLEs could use different algorithms. It could also be a "broken cadence" problem. See this Grass Valley document "Understanding Cadence" starting at page 35 for examples of conversion algorithms:
  wwwapps.grassvalley.com/docs/Application...adence_v4.0.0.20.pdf

To check this  your colorist could import one clip (which is apparently 59.94) to a new 23.98 Resolve project, keep it at 23.98 (IOW don't allow Resolve to change it), do a simple but obvious color correction, then export that one clip using Resolve's FCPX preset, then he loads that into FCPX and stacks it on the original clip using 50% opacity. As previously discussed using that preset Resolve should export the clip at the original frame rate, regardless of the project frame rate. The reason for using the 23.98 project is to match the previously-stated project frame rate.

If he sees no difference then it's possibly something else he did to the main project when grading it. If he does see a difference on the simple one-clip project, then other possibilities would need investigation.

Please Log in or Create an account to join the conversation.

Last edit: by joema.
  • Page:
  • 1