Step away from the Apple Keyer now! (And build a better one of your own.)

[Simon Ubsdell]

[Jeremy Garchow] “Of course I’d fix it, and it would keep me up at night.

Maybe Apple doesn’t care becuase you, Bret, and I are the only people who do. Everyone else seems pleased.

So, what does that say about us??? ;)”

Haha. Yes, it’s worrying to be the kind of person that worries about such stuff 😉

Won’t get us a job at Apple, though.

Simon Ubsdell
http://www.tokyo-uk.com

[Shawn Miller]

[Simon Ubsdell] “Just to come back round to your point about beta-testing – beta-testing is pretty useless for picking up stuff like this unless you’re very lucky. You simply have to have systems in place to pick up basic technical flaws – even if it’s just your co-workers taking sufficient interest in what you’re doing.

So I don’t blame Apple’s scanty beta testing program – I blame the lack of pride in their job of the people doing the work, sorry to say.”

I like your scenario Simon, and I completely agree that it’s plausible. Let me add a wrinkle to your thought process… from a software development standpoint, there may be absolutely nothing wrong the keyer! I think it’s just as possible that edge refinement wasn’t part of the spec to begin with… so, not only did Joe (or Josephine) do his/her job, but they probably slept just fine knowing that they built the keyer ‘to spec’. In that case, there would be nothing for QA/test to catch; does the keyer crash – no, is it fast – yes, does it work as expected, yes – pass.

It’s possible that the product team doesn’t think that editors need that sort of complexity or control. So, this really isn’t a technical flaw, but a feature found in most other keyers… just not this one.

Shawn

[Simon Ubsdell]

[Shawn Miller] “I think it’s just as possible that edge refinement wasn’t part of the spec to begin with… “

Sorry, Shawn – I think I may have not made it clear what I was talking about here. This is looping back to an earlier thread where we were talking about how the lightwrap makes a complete mess of the edges of your image because the blur operation hasn’t been set up correctly:

https://forums.creativecow.net/readpost/335/64474

Here’s the screen grab I posted again to make it clear what we were talking about:

So, nothing as “complicated” or demanding as edge control.

Just a complete hash made of a feature they did decide to implement.

Simon Ubsdell
http://www.tokyo-uk.com

[Shawn Miller]

[Simon Ubsdell] “[Shawn Miller] “I think it’s just as possible that edge refinement wasn’t part of the spec to begin with… ”

Sorry, Shawn – I think I may have not made it clear what I was talking about here. This is looping back to an earlier thread where we were talking about how the lightwrap makes a complete mess of the edges of your image because the blur operation hasn’t been set up correctly:

https://forums.creativecow.net/readpost/335/64474

Here’s the screen grab I posted again to make it clear what we were talking about:”

Ah, I see. I did misunderstand, thanks for the correction. Reading through the thread, I think you’re right. This does seem like a case of, ‘it’s good enough, they’ll never notice’ product development.

Shawn

[Walter Soyka]

[Simon Ubsdell] “No-one at Apple pro Apps was interested enough in their own keying product to spend their own time playing with it and discover this elementary mistake. Really?? No-one at Apple actually uses this stuff? Yes. That scenario I can believe.”

Sad, but probably true.

[Simon Ubsdell] “Just to come back round to your point about beta-testing – beta-testing is pretty useless for picking up stuff like this unless you’re very lucky. You simply have to have systems in place to pick up basic technical flaws – even if it’s just your co-workers taking sufficient interest in what you’re doing. So I don’t blame Apple’s scanty beta testing program – I blame the lack of pride in their job of the people doing the work, sorry to say.”

I disagree a bit here. I’ve been on beta teams for a range of different developers and products, and the problems with the keyer are exactly the sort of thing that should get picked up right away in real world projects.

You’re right that I should not assume that the Apple beta program missed it. In fact, I am sure that if anyone with keying experience testing actually tried this, they ran into problems. As I talked about above, I ran into problems on the second shot I ever tried the keyer with, and ultimately found it acceptable for only a single shot in the entire project. It’s not corner cases where the keyer fails. It’s corner cases where it succeeds! If you have anyone testing the keyer in the real world, these problems should be obvious.

Here’s how that conversation would go:

Devs:
“Hey beta team, we’ve got a brand new chroma keyer for M5 and FCPX. It’s super cool because it guesses the screen color and has groovy on-screen controls for edge manipulation. Give it a try and let us know what you think.”

Testers:
“Hey dev team, great-looking feature. I always hated using the color picker to find the screen color. Those OSCs are really intuitive. And wowsers, this is fast! But hey, by the way, you should know that we tried this on a few shots and we can’t get the keyer to give us usable edges. Oh, and the light wrap seems to be wrongly affecting frame boundaries. I’m opening tickets for both issues. Should I send sample footage?”

If this wasn’t caught by a tester, Apple is operating a near-useless beta program. But if this was caught by a tester, then Apple is compounding ignorance during development with apathy to real-world testing.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Simon Ubsdell]

[Walter Soyka] “I disagree a bit here. I’ve been on beta teams for a range of different developers and products, and the problems with the keyer are exactly the sort of thing that should get picked up right away in real world projects. “

I’d say that you are a pretty abnormally perceptive, diligent and talented beta tester so your experience might not be typical 😉

[Walter Soyka] “It’s not corner cases where the keyer fails. It’s corner cases where it succeeds! If you have anyone testing the keyer in the real world, these problems should be obvious.”

That’s pretty harsh! But fair. What really amazes me is that it even went out to beta testing in this form. Does no-one on the development team know how to evaluate this stuff properly?

[Walter Soyka] “If this wasn’t caught by a tester, Apple is operating a near-useless beta program. But if this was caught by a tester, then Apple is compounding ignorance during development with apathy to real-world testing.”

That’s really harsh! But it’s very hard to sketch out any scenario that justifies what we’re seeing here.

I’d be interested to see what you think about how this reflects on the rest of Apple’s image processing offerings for pro video. What does it say about them that they either let this stuff go, or (worse?) they aren’t able to detect that it’s no good enough in the first place?

There are certainly a few obvious Core Image units that are pretty iffy, to say the least …

Simon Ubsdell
http://www.tokyo-uk.com

[Walter Soyka]

[Simon Ubsdell] “That’s really harsh! But it’s very hard to sketch out any scenario that justifies what we’re seeing here.”

Maybe I am being too harsh. As Jeremy pointed out elsewhere, we are at talking about an NLE.

Also, we will never know all the positive impacts the beta program had on development. It’s not fair of me to lay a questionable keyer at their feet.

But then again, this all really comes from Motion which is a compositor, not an NLE, and the chroma keyer was touted as a major feature, and it simply fails a lot. There’s no getting around the fact that a lot of people should have recognized this, and no one has remedied it in the last 2.5 years.

Really, while you are upset that Apple could release a shoddy keyer, I’m probably more upset that we all as a community bought into the hype that it was great. I am embarrassed by the quote I pulled from myself a year ago, because I saw these problems with my own two eyes, but in a total Emperor’s New Clothes moment, I thought *I* was the problem.

So in other words, not only did Apple release a shoddy keyer, they got a free pass on it from us, their customers, for the last 30 months. More than a free pass — glowing recommendations. I wish we all had been more discerning. It would make the product better.

[Simon Ubsdell] “I’d be interested to see what you think about how this reflects on the rest of Apple’s image processing offerings for pro video. What does it say about them that they either let this stuff go, or (worse?) they aren’t able to detect that it’s no good enough in the first place? There are certainly a few obvious Core Image units that are pretty iffy, to say the least …”

I think the fact that Apple has an offering like Core Image is probably more important than whether it’s good. This might sound familiar, but I see it as a case of high floor, low ceiling. The developers who want something quick and easy will use it, but the developers who really care about image processing will write their own anyway.

What are you seeing in Core Image that you don’t like?

Personally, I’ve been kind of drifting away from the Mac platform over the last couple years, in part because I’ve been using PCs a lot more and in part because image processing on the Mac, after a burst of promising development and momentum, just seemed to stop.

I loved Quartz Composer, but it’s absolutely languishing with an uncertain future while all the good stuff is coming exclusively from third-party developers. Now I’m supporting Vuo [link]. I’ve done some work with Processing. I’m starting to geek out with GLSL and TouchDesigner. There is a lot of cool stuff happening elsewhere.

All that said, I’m starting a new project in Motion today.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Simon Ubsdell]

[Walter Soyka] “I think the fact that Apple has an offering like Core Image is probably more important than whether it’s good. This might sound familiar, but I see it as a case of high floor, low ceiling. The developers who want something quick and easy will use it, but the developers who really care about image processing will write their own anyway.”

Yes, indeed, there’s nothing wrong with Core Image itself. My point was rather about the quality of some of the units that Apple built and which form an intrinsic part of the architecture of both Motion and FCP X.

Writing your own Core Image kernels is definitely the way to go. In fact I’ve spent an interesting afternoon speculatively building my ideal (Color Difference) keyer in Quartz Composer and as so often I found it’s easier and better to write your own custom kernels.

It’s interesting to hear you’re getting into Vuo – I’ve keeping a watchful eye and it looks very promising.

Simon Ubsdell
http://www.tokyo-uk.com

[Chris Wright]

hey, thanks for the vid. you inspired me to make an after effects procedural keyer. It has the same edge detail as keylight plus more refining tools. i made it open source so that you can port it into plugins.

aea cs5.5 aep template
https://f1.creativecow.net/7088/7088