[Michael Mccune]

Clarification: when Fusion 9 fails to save the composition, clicking the “Save” button at the bottom of the “Save” dialog does nothing.

That is, the dialog box remains and is not dismissed, as is the case when the composition is successfully saved.

Thanks, Mike

[Michael Mccune]

ANSWER: the good news is that Fusion 9 does seem to open the DNx formats successfully; that is, without the error message, “Failed to load …”

The problem seems to have been a conflict between Fusion 8 and Fusion 9.

Found an unexpected “Fusion 8” tag on the opening splash screen; perhaps the old icon was still opening version 8.

Uninstalled Fusion 8 and–so far–the DNx variants load as expected, with no error messages.

Thanks, Mike

[Michael Mccune]

Right. Tried that. Same failure to launch.

In fact, watching Microsoft Process Explorer showing interesting–and unexpected–processes running even while uninstalling it: a Resolve process continued to run until I restarted the PC (!).

Gotta be the “Beta Blues” song that I am singing.

No doubt this stuff is incredibly complicated under the hood.

Surely in a week BM will offer updates. Looking forward to the many helpful feature additions.

Thanks, Mike

[Michael Mccune]

Thanks again, Simon.

Yes, DNx of all flavors work in Resolve, both in and out on my PC.

And the “anomaly” I referred to was simply the fact that only the compressed codecs, the ones that do not open in Fusion, showed 24 fps, not the 29.970 fps of Resolve and the Resolve timeline.

Question: do any compressed codecs import into Fusion, in your experience???

Also, just for my curiosity, please describe your hardware and the typical render fps that you consider normal or adequate. 5 fps??? 25 fps??? 75 fps??? I realize that the node complexity will impact this. But the idea is to understand the general level of render performance is normal or acceptable for your work.

I have the impression that professionals usually run at full-tilt: that is, multiple graphics cards (Quadro or other series???), multiple processors, as many CPU cores as possible and the full complement of ram.

Today I asked Nvidia about any CUDA or OpenCL configuration tips for Resolve and Fusion and they said–and would only say–that GeForce 1080 series cards are not recommended, use Quadro. Just for due diligence, however, one has to recollect that there is almost no comparative benchmarking or other specific info.

One description from an Nvidia engineer a few years ago seemed simply to highlight driver stability rather than performance. This seems to mean that it will never crash under demanding commercial graphics applications, but otherwise they make no claims even to parity regarding performance.

Thanks again, Simon.
Mike

[Michael Mccune]

Thanks for the quick reply, Simon.

Yes, doing this work on a PC.

And the Resolve original source video is 29.97 fps and the Resolve time line is 29.97, as it is shown in the Project Settings, the Edit tab media pool info and in the Media tab clip metadata. So that does actually seem to be 29.97.

But the fps for the compressed codecs may not matter, as we see that this anomaly is seen in the failed compressed codecs only, not the successful frame based codecs.

============
FWIW, here is a list of the codecs tested, seeming to show that frame based codecs work but not the compressed codecs.

Love to get a DnX codec to work in Fusion, such as DnX HD 220 10-bit.

1) 1 min QT DnX HD == 1.54 GB; 44 sec encode time == fails in Fusion; _1_min_QT_DnX_HD_v01 cannot get Parameter for Input at time 0 _1_min_QT_DnX_HD_v01 failed at time 0

2) 1 min QT 444 == 3.09 GB; 60 sec encode time == fails in Fusion; V10001_1_min_QT_444 failed at time 0 _1_min_QT_444_v01 cannot get Parameter for Input at time 0_1_min_QT_444_v01 failed at time 0

3) 1 min QT Uncom RGB == 10.4 GB; 55 sec encode time == OK in Fusion

4) 1 min TIFF RGB 8bit == 10.4 GB; 148 sec sec encode time == OK in Fusion

5) 1 min DPX RGB 10bit == 14 GB; 109 sec sec encode time == OK in Fusion
6) 1 min DPX RGBA == 14 GB; 193 sec sec encode time == OK in Fusion

============

Thanks for your thoughts on this, Simon.
Mike

[Michael Mccune]

Simon; thanks for the reply and many thanks again for your other tutorials, which–unlike most tutorials out there– I have found to be even more rewarding with multiple viewings. They seem to have so many helpful pointers plus a clear overall plan for your discussions.

Yes, the OpenCL setting improved things markedly.

It had been running about 5fps, and by rendering the RGB channels as b/w, the rate went to about 9 fps and the TIFF frame size went from about 3.7 MB to about 1.8 MB, which seems about right. Then using OpenCL the rate went to about 18-19 fps. Other clips seem slower, perhaps 8-10-12 fps.

The GPU utilization still is only about 18%-20%.

Using SSDs for source and record drives and 6 cores x2, but only 32 GB ram. And it is all used. Max ram set aside in Preferences.

Regarding Fusion Connect, I am delighted that you seem to have confidence in its capabilities. That is, for this matte generation task specifically. Of course it returns to Resolve a flat file. This of course loses the Resolve graded node tree.

So, am I understanding this correctly? That is, this is really intended to be a one-shot technique?

There must be a work-around to allow one to revisit the original Resolve node grade, as one can return to the Fusion grade. But in Resolve, too, there may be requested changes.

Perhaps append a gallery grade from a Resolve still into a new clip copy, a replacement copy, of the timeline clip (which had become a Fusion timeline clip)???

Likewise, perhaps preserve the Fusion Connect node tree (copy paste into a temp Fusion project) while still in Fusion. Then after color changes as needed are applied in Resolve to a clean copy of the timeline clip with the original gallery grade applied as a starting point. Create a new Fusion Connect clip and paste the previous Fusion node tree into the new Fusion Connect clip. Make changes in Fusion as needed. Then render out again to the Resolve timeline.

What do you think???

Convoluted. Gotta be a better way.

Thanks again, Mike

[Michael Mccune]

Thanks, Glenn; you are right I believe about the number: 29.97 vs. 29.976.

BUT THE SOLUTION TO THE MYSTERY OF THE EXTERNAL MATTE FRAME OFFSET IS AT THE END OF THIS POST!

There are many online references to 29.976 but there is no derivation for that which I could find in a brief search today. So I agree and lets stick with 29.97.

As you must know, the 29.97 number comes from 30/1.001, which harks back to the first color TV days and their need to shift the frequency harmonics slightly. I understand that to reduce all the phasing from the color subcarrier, etc., they had to calculate out several orders of harmonics. Amazing. And they did all the math without computers.

Anyway, here is the actual solution, to the external matte offset problem, gleaned from this tutorial, https://youtu.be/zcUtbdT5Uxo, which mentions that imported external mattes line up with the start of the original clip, presumably the source time code.

And, indeed, my test clip had been shortened at the head by about 37 frames.

So, if one exports a portion of a clip starting at some point into the clip, not the start of the clip, and then the Fusion matte made from that portion is imported back into Resolve as an external matte it will not line up, as expected, with the start of the exported portion of the entire clip. Rather the external matte will line up with the start of the entire clip prior to export.

Why??? Dunno. Something that could be clarified in a later version of Resolve, sez I.

But how to fix it??? ANSWER: The external matte key tab includes an offset option. This critter is difficult to manage due to very high sensitivity: slight mouse movements can translate to relatively huge frame increments or decrements. And there is no nudge keyboard option that I know of.

With some fiddling with this control, the -37 frames adjustment did offset the matte to match the original video.

Funny but I tried this option before but gave up, supposing I didn’t understand how to use it as it would jump wildly in the frame count even with tiny mouse movements. And of course I needed only about 37 frames. (Actually, I wonder if the huge offset range results from scaling the long length of the entire original clip: not hard to jump to an offset with hundreds of thousands of frames. But this behavior makes small changes of a few frames tricky and actually difficult.

So I guess Resolve might include this as a feature request and refine this option to provide some sort of scaling and nudge capability.

Thanks again, for the commiseration, Glenn. Mike

[Michael Mccune]

Glenn; I was hoping you would chime in. Yes, trouble with mattes.

And, yes, all the work, from source to final, is 29.976.

You asked about my Fusion frame rate setting. So let’s start with my confusion over setting the Fusion frame rate:

Just for fun, try a search for, “29.976” in the Fusion User Manual or the Fusion Tool Manual (try Ctl+F to bring up a local search box for the document). What do you find??? There are NO entries in either manual for the term, “29.97” or for, “29.976.”

First, we note that Fusion has two “frame format” setting locations within Preferences, under Global and Default Settings and under and Composition/General.

These frame format pages each have frame rate data entry fields but they are pre-populated only with whole number integer frame rates, such as, “24, 25, 26, 27, 28, 29, 30, 31,” etc. Again, no “29.976 option.

But it seems that it is possible manually to enter a fractional frame rate in these fields. But not really.

That is, the frame rate field entry is constrained to two decimal digits, e.g., “29.97.” We know the actual frame rate is, ‘29.976.” So it is not possible, either by choosing an item in the drop-down list or by entering, “29.976” in this field.

So how is one supposed to understand this frame rate setting business in Fusion???? Seriously, what is the logic???

FWIW, the frame rate–so says the documentation–is determined by the frame format which is resolution independent. Don’t quite follow the logic, however.

Or, try a search for, “frame rate:” there is ONE entry in the Tool Reference but it only relates to stretching duration. In the User Manual it says, “Since Fusion is
resolution independent, comps do not ask you to define a resolution or frame rate. Those settings are determined in the Globals [sic] Frame Format preferences,” (from page 25). Don’t understand how resolution–to me that means the total number of pixels, horizontal x vertical, in some HxV ratio that is the format— can yield a rate. That is, a given format, say, 1920×1080, can be recorded at widely variable frame rates.

So how to they get there, a derived frame rate???

(Somewhere in this mix is also the question that the tiff file format does not have, I understand, ANY integral information about frame rate, as might be embedded in other codecs and file formats.)

If you find you have any helpful corrections or thoughts on this, Glenn, glad to hear them!

Thanks, Mike

[Michael Mccune]

Tero; thanks for the reply.

Perhaps I was not clear about the fact that the same Fusion clip is imported into Resolve both as a video clip and as a matte.

In the case of the clip imported as video and placed on a track above the original source track, the time code matches at head and tails, so perfectly in sync.

But in the case of the import as an external matte however the time codes–and the visual matte–do not match and are visually out of sync.

So where would the issue of a frame rate difference enter into the process???

Thanks, Mike

[Michael Mccune]

Thanks, Simon; I had wondered about the frame setting and various sizing options such as transform on the output.

Your tutorials are a favorite: clear, articulate, and well done.

Thanks, Mike