FORUMS: list search recent posts

The theory and mechanics of rendering in Final Cut Pro

COW Forums : Apple Final Cut Pro Legacy

<< PREVIOUS   •   FAQ   •   VIEW ALL   •   PRINT   •   NEXT >>
Miranda Raimondi
The theory and mechanics of rendering in Final Cut Pro
on Mar 3, 2011 at 1:23:54 am

What is the theory and what are the mechanics of how Final Cut Pro, and the Apple platform, renders graphics and video.

aka) -How- does Final Cut Pro do its rendering?

Thank you,
Miranda

Samples of my collaborative work are available for viewing at Vimeo.com. Search 'Miranda Raimondi' or 'Samuel Pellman'.


Return to posts index

Rafael Amador
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 3, 2011 at 1:43:49 am

Hi Miranda,
In FC you ca render in 8 or 10b (Floating point) on YUV and in merely 8b in RGB.
You set that on your "Sequence Settings".
If you put graphics in a YUV sequence, those are converted to YUV before being rendered in that color space.
rafael

http://www.nagavideo.com


Return to posts index

Andy Mees
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 3, 2011 at 4:59:26 am

Wow, that question is a little broad Miranda. What is it that you actually want or need to know? At the lower level ie the mecahnics of rendering in FCP, bear in mind that no one here is likely to be privy to the FCP code base ... and even if they were they are almost certainly not authorized to reveal it for all the world to see on a public forum. But very broadly, when any frame of the source does not match the properties of the target (and this can be because your source and target properties simply don't match or because you are modifying the source to suit your purpose) then FCP must create a new frame from that source according to the modifications applied and the properties of the target. When the calculations required to create each new frame cannot be made faster than the target frame rate, then you need to "render" ... in which case the new frames are pre-calculated and written to disk as a physical media file (a render file), just like the original source clip, and thereafter referenced by FCP in place of that original media.


Return to posts index


Kent Jakusz
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 3, 2011 at 4:48:27 pm

Thanks Andy. Well said.
I frequently have to re render the same clips. For example, I am currently working on a 37 minute ProRes 422 1920 x 1080/29.97 timeline. In this timeline there are 8 clips of (guess) 300 that will not stay rendered. Every hour or so they require re rendering. Any suggestions?

All clips originated in HDSLR's, all have been converted from H.264 to ProRes via MPEG, all 300 clip properties are identical in the Browser. There doesn't seem to be any common attribute among these clips.
Puzzled
Kent


Return to posts index

Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 3, 2011 at 8:50:42 pm

Hello Rafael, Andy and Kent,

Thank you for your timely and thoughtful replies.

Yes, the question is broad, and I suppose that I was wanting to get as close to the source code as possible without actually asking someone to reveal FCP's source code.

The question, perhaps, should have been accompanied with some context for why, exactly, I am asking about the theory and mechanics of rendering in FCP. That context would have been, and eventually will be, submitted as a targetable question, that has real world applications to an issue that I am encountering with some of my editing.

Briefly stated, the footage that I am working with has some graphic attributes that can sometimes be seen with the still and moving playhead in both the viewer and the canvas windows. I say sometimes because rendering the footage, I guess -and this is where I begin to theorize-, 'recompresses?' the footage in such a way so as to banish those graphic attributes which are the very object of what I want to work with and display in the rendered, finished piece.

I would understand more about what was going on if rendering was the only action that hid or banished, overwrote or counteracted those graphic attributes.

However, whenever I view FCP's sequence properties or sequence settings it is as if final cut pro gets 'reminded' to 'reinterpret' the footage in such a way as to also hide those coveted graphic attributes. Sometimes those attributes go m.i.a. without any user prompting, which is even more perplexing to me, but which also tells me that the disappearing act may have something to do with how many computing and processing resources are available to FCP at any point in time.

I ask the question about rendering because I want to know whether FCP is somehow recompressing or rewriting frames as if it were rendering those frames for real-time playback purposes. If so, what would it be, about that real-time progression, or any playback, whether timebound or still, that would hide such graphic attributes?

Would it be the interaction between playback frame rate and interlacing? But that would not explain why still frames would seize to show those attributes.
Does the scenario that I describe have to do with sampling rates of Luma, Color and Alpha channels?
With refreshed, or more active, gamma corrections?
Is the alpha or luma channel simply reinterpreted differently before and after the graphic attributes go m.i.a.?
Does the scenario that I describe have to do with the render controls, ie) frame blending?
Does the scenario that I describe have to do with yet other sequence setting specifications, like the codec that is chosen? The field dominance that is employed? The sequence's, the codec's and the original source material's editing timebase?

I am trying to identify the cause of the graphic attribute's disappearance. I ask these questions about rendering because I wonder whether a new frame is being generated from the source material as if FCP was pre-rendering the material for still or moving playback purposes.

I want to know which calculation that FCP is performing (with and without prompting) that is responsible for the graphic attribute's disappearance. What sort of refresh is occurring to make that graphic attribute disappear?

Though I have thoroughly described the context for my question about the theory and mechanics of rendering, another question still stands:

Why, ultimately, would there be visual disparities between rendered and unrendered frames even if those new, rendered frames were recompressed versions of the unrendered frames. Shouldn't such frames be translated with 100% fidelity? Or are pixels, the pixels that I want, dropped in that recompression or refresh? Could I avoid this by working in an uncompressed codec?

I will also compose and post a follow-up inquiry for the specific scenario that I describe. That follow-up inquiry will appear as another thread and, if approved, will be searchable with the following subject line: "Graphic attribute's unprompted disappearing act'. Please feel free to post your answers in either thread if you have any inklings as to what may be going on.

Thank you so much,
Miranda


Return to posts index

Andy Mees
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 4, 2011 at 1:25:55 am

Hey Mirada

[Miranda Raimondi] "The question, perhaps, should have been accompanied with some context "

Yes please ... always a good idea.

[Miranda Raimondi] "Briefly stated, "

Uh oh ... we need copious and specific details, not brief statements ;-)

[Miranda Raimondi] "The footage that I am working with has some graphic attributes... "

Vague! Details Miranda, we need details. :-) Ok, lets go back to the basics.
What are the format properties for these graphics? Select one and press Cmd-9. Write down (or take a screenshot of) what you see.
What are the sequence properties? With sequence active, press Cmd zero. Write down (or take a screenshot of) what you see.
What are the RT plaback properties of the sequence? Click on the RT button top left of the Timeline window and note down all the checked settings.
Post back with all those details.
And how are you viewing all this ie on what image are you basing the critical assessment? Are you looking in the Canvas window, and if so, at what scale? The Canvas window is just a proxy of the full image, it's not necessarily showing you everything. For example, at any resolution other than 100% will only display the Upper Field of an interlaced image ... so if you are using it for monitoring of interlaced footage then you'd best set it's display scale to 100%. For critical assessment tho, what you should really be looking at is a full resolution output of your image on an appropriate reference monitor (attached to your system by an I/O device). If you are assessing interlaced material then that reference monitor needs to be capable of displaying that interlaced image properly.

[Miranda Raimondi] "I ask the question about rendering because I want to know whether FCP is somehow recompressing or rewriting frames as if it were rendering those frames for real-time playback purposes"

Yes it is.

[Miranda Raimondi] "If so, what would it be, about that real-time progression, or any playback, whether timebound or still, that would hide such graphic attributes? Would it be the interaction between playback frame rate and interlacing? But that would not explain why still frames would seize to show those attributes."

I already mentioned that the Canvas window at any resolution other than 100% will only display the Upper Field of an interlaced image ... if you have a detail in your source image that when rendered to the sequence properties only takes up 1 scan line, and that scan line happens to lie be an odd number (Upper field) then you won't see it in the Canvas window at anything other than 100% scale ... doesn't matter if the playhead is moving or still, its the same. On an appropriate reference monitor you would see the image as it should be but that same element will appear to flicker as it is switched on and off with the sequential display of odd and even frames. Have the graphic remade more appropriate attention to the target settings ... or just add a tiny amount of blur to it in your sequence, just enough to spread that detail onto a second scan line ... that should resolve the issue.

Cheers
Andy


Return to posts index


Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 4, 2011 at 6:24:30 pm

Hi Andy,

Thank you for your patience:

The details about the RT rating, the sequence settings and the format settings are available in the following pdf attached in the file below:

1705_pursuingthedesiredgraphicattributes.pdf.zip

The pdf includes the visual results of viewing the graphic in canvas window that displays at a setting of 100%, 99% and 400%. All views successfully display the graphic (as seen as pixelated and sometimes interlaced black and grey dots) - those desired graphical attributes are chroma key artifacts.

Thank you for your comments about proper viewing of an image on an output monitor or I/O device. Unfortunately I do not have that capability at this time.

Thanks again,
Miranda

Miranda Raimondi

Samples of my collaborative work with Dr. Samuel Pellman are available for viewing at http://www.vimeo.com.

Please search 'Samuel Pellman' or 'Miranda Raimondi' in vimeo.com's search function in order to view those samples.


Return to posts index

Rafael Amador
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 5, 2011 at 6:01:51 am

Hi Miranda,
I see two things that I don't like:
- Your graphics are ".dv".
- They are interlaced.
That is enough to complicate things, but things may be further complicated if your sequence is DV NTSC.
I would start by changing the sequence codec to something better (Prores) and have a look to see if things improve on the Canvas..
rafael

http://www.nagavideo.com


Return to posts index

Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 6, 2011 at 3:01:36 am

Hi Rafael,

Thank you for your reply.

Yes, the source material is dv material, and it is interlaced. Though I can address this with my collaborator in the future, for now, I really should continue to work with the material that I have.

I am coming across many strange, and interesting, though frustrating, idiosyncrasies with the source material that I am working with.

For instance, when I alter the speed of the dv tracks to 33%, the desired graphic attributes remain in the canvas and can be displayed and played back even after the dv track has been rendered.

Nevertheless it is as if rendering 'flattens' those desired graphic attributes. Whereas more than one field (and the graphic attributes in those fields) were visible prior to rendering, rendering seems to reduce those fields, and the graphic attributes within those fields, into a single 2-dimensional layer. In other words, prior to rendering, the desired graphic attributes were being displayed in a kind of 3-dimensional layering that was very much akin to the kind of human perception of superimposition that occurs when two objects occupy the same line of vision but are separated in space, thus making the object that is closer to the observer, in z-space, appear to occlude the object that is farther from the viewer, in z-space. Upon being rendered the image seems to lose both its 3-dimensionality and the independence of its composite 'objects', which I have been describing as 'graphic attributes'. After rendering, those objects appear to occupy the same z-space.

As I also mentioned in the attached pdf document, merely opening the sequence attributes seems to cause final cut pro to reinterpret the source material as if it were rendering the source material. Strange.

Thank you in advance,
I will continue to learn more about what is going on.
I will update you of my findings provided that you are still interested.

Miranda

Miranda Raimondi

Samples of my collaborative work with Dr. Samuel Pellman are available for viewing at http://www.vimeo.com.

Please search 'Samuel Pellman' or 'Miranda Raimondi' in vimeo.com's search function in order to view those samples.


Return to posts index


Andy Mees
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 6, 2011 at 5:51:45 am

[Miranda Raimondi] "Though I can address this with my collaborator in the future, for now, I really should continue to work with the material that I have."

I'm sorry Miranda but I'm not at all sure that this is wise. Yes, you should certainly address this with your collaborator in the future, as between the two of you, you seem to have invented a Frankenstein format which is not working well. According to the info you posted in the earlier document, the Compressor (codec) of your footage is "None" with a data rate of 39.6 MB/s, but the filename extension is clearly ".dv" ... which seems to suggest you have somehow wrapped the footage up and are presenting it to FCP as a DV transport stream. Not good. I don't quite know how or why you managed to arrive at this point but my advice would be to fix this before you dig yourself in any deeper. Rewrap your footage as a regular Quicktime Movie file, and at the same time you would do well to reencode it using a compressor that is more fully supported for "RT" in Final Cut Pro ... I would recommend you choose Apple ProRes 422.

[Miranda Raimondi] "After rendering, those objects appear to occupy the same z-space."

I would suggest that that is entirely as one would expect ... you are after all (for the time being at least) working in a 2D medium. If you have multiple items being composited in a single scene, the act of rendering does indeed "flatten" them.

[Miranda Raimondi] "As I also mentioned in the attached pdf document, merely opening the sequence attributes seems to cause final cut pro to reinterpret the source material as if it were rendering the source material. Strange."

Yes, that one certainly is rather bizarre. I'm guessing that your system is struggling with this strange footage and that the RT engine is therefore dropping to a lower resolution which is causing elements of your footage to appear to disappear during some playback operation. I would hesitantly suggest that you need to understand that FCP employs more than one potential rendering path (software or hardware), and that it also employs more than one potential display path, being RT and non-RT. I say "hesitantly" only because this is not at all the kind of understanding that any normal user of the software should want or need to understand. If you are work with supported formats, then the software will work exactly as you should expect it to work. If not ....

But I digress. You're not coming across at all as regular user at all, and I believe I've already made the point that you should fix your workflow, not go on a wild quest to understand the inner workings of FCP ... but yet much of that inner workings info is no secret so I'll cut to the chase:

The RT Display Path

During normal operation, the RT engine is responsible for updating the video displayed in the canvas window and on any external monitor. The RT engine is engaged during playback and during “scrubbing.” In some circumstances, sequences cannot play back in realtime on all machines. When this happens, the RT engine can adjust quality in several ways in order to maintain reasonable playback performance (Low/Medium/Dynamic).

The Non-RT Display Path

In addition to the RT engine, Final Cut has another display path which is sometimes used to draw video to the canvas window and external devices. This display path predates the RT engine, and is used for certain display options which aren’t supported by the RT display path. Unlike the RT engine, the Non-RT display path does not display the same image in both the canvas window and the external monitor. When updating a frame, the Non-RT engine renders an image for display in the window, and then renders another image later to send to the external monitor. The external monitor is updated at idle time, so if you are scrubbing quickly you may notice that the canvas window updates more frequently than the external device.

One of the things that forces a clip down the non-RT display path is having Show Excess Luma or Chroma enabled ... try enabling that (press Control Z) and see if that changes what you see in the Canvas window. Then again, that won't work because if your format. "None" is not supported for the RT display path.

Hope we're getting somewhere
Andy


Return to posts index

Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 8, 2011 at 8:56:12 pm

Hi Andy,

Thank you so much for your timely reply. I apologize for not being able to offer the same courtesy... my internet has been amiss, or remiss.

Much in keeping with the theme of this thread, that theme being related to the relative normality of FCP's inner and outer workings, and those who preside by those workings, I must confess that I am no stranger to playing the 'abbie' in the 'ab'normal of Frankesteinian screen plays. Those who are still working in 2d mediums are still, after all, playing with magic lanterns and negotiating the frontiers of perception and illusion.

But I digress...

Back to FCP and to those graphic attributes:

Your advice is well taken and many of your assessments are, I believe, right on target. I do have to wonder whether, rather than dropping down to a lower resolution, those graphic attributes, and their lack thereof, aren't a result of the way that FCP blends frames according to different rendering and playback pathways.

It would seem entirely plausible to me that accessing the sequence's sequence settings might send FCP's frame blending (and possibly other rendering and playback) functions down another (and possibly more default or preset) pathway where processes may, for example, revert from software computational channels to hardware computational channels, and where those graphic attributes are, in this example, results of how FCP's software interpolates frame blending as opposed to how FCP's hardware handles frame blending.

While the hypothesis that drops in resolution are responsible for FCP's display or non-display of those graphic attributes is quite a mature guess, I would hesitate to support that hypothesis given that experiments with FCP's de-interlacing functions (including that of shifting fields, and the various degrees of flicker filter that are available, and even the various blurs that are available do nothing to either hide (when the attributes are in view) or reveal (when the attributes are absent from view) those graphic attributes.

There seems to be an all or nothing kind of shift to the graphic attributes' display, which would suggest to me that a 'processing', as opposed to an 'aesthetic', shift is occurring and is determining what is and is not displayed on the computer monitor irregardless of whether the material has been rendered or not, or whether the material is actively being played back or not, no less in real time (RT) or not. Those graphic attributes, I've found, are sometimes (but not always) capable of prevailing even after being rendered, yet they taken on a different caliber. They lose their 3 dimensionality pre- and post-rendering, which, as you say, is to be expected (holograms are not exactly pertinent to this specific question, yet).

Thanks again,
Cheers,

Miranda

Oh, and well put, about those attributes 'appearing to disappear'. I like that.
Houdini surely makes a cameo in this Frankenstein mix... and maybe Watson too (the late, of course).

Miranda Raimondi

Samples of my collaborative work with Dr. Samuel Pellman are available for viewing at http://www.vimeo.com.

Please search 'Samuel Pellman' or 'Miranda Raimondi' in vimeo.com's search function in order to view those samples.


Return to posts index

Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 8, 2011 at 9:07:49 pm

Yes, we are getting somewhere.


Return to posts index

Rafael Amador
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 9, 2011 at 1:51:17 am

Hi Miranda,
".dv" (DV Sequence), IS NOT SUPPORTED by FC. Always needs to be rendered, and as Andy suggests, the file will be flattened to before processed by FC.
That shouldn't be on your sequence.
I would start by trying to re-wrap that stuff again as QT.
You may try MPGStreamclip: "Save as > QT".
No rendering required and you may get back the graphics with the proper attributes on a proper QT file.
rafael

http://www.nagavideo.com


Return to posts index

Miranda Raimondi
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 9, 2011 at 1:57:23 am

Thank you Rafael. Thank you to all who contributed to this thread.

Miranda Raimondi

Samples of my collaborative work with Dr. Samuel Pellman are available for viewing at http://www.vimeo.com.

Please search 'Samuel Pellman' or 'Miranda Raimondi' in vimeo.com's search function in order to view those samples.


Return to posts index

James Mortner
Re: The theory and mechanics of rendering in Final Cut Pro
on Mar 9, 2011 at 2:07:57 pm

Very interesting thread everyone !

Just to add, don't the black dots on the edge of the doecahedron graphic in the PDF look like alpha masks gone a bit wrong ?

If you select the clip in the timeline and press "cmd_9" again, try changing the alpha properties by right-clicking on the word "straight" next to "alpha" and selecting black, white or ignore. This might change the way the dv clip composites and renders.

But i would still find a better format like the others said.

Cheers !


Return to posts index

<< PREVIOUS   •   VIEW ALL   •   PRINT   •   NEXT >>
© 2018 CreativeCOW.net All Rights Reserved
[TOP]