-
Maybe 10 Years is Enough for Final Cut Pro X – Philip Hodgetts
Maybe 10 Years is Enough for Final Cut Pro X – Philip Hodgetts
https://www.philiphodgetts.com/2018/06/maybe-10-years-is-enough-for-final-cut-pro-x/I had been thinking about the “next direction” given that Apple did say there was a ten year plan with FCPX. Given how far into that we are at this point, I thought that maybe just as FCPX is settling, much like FCP legacy, we are actually nearing its final leg. Sometime around 2021 the Apple NLE ecosystem may tremble again. Perhaps it will be an NLE built around machine learning.
-
15 Replies
-
I just saw this a couple of days ago –
https://www.fcp.co/final-cut-pro/news/2094-automatically-put-markers-on-the-beat-with-beatmark-x-for-final-cut-pro-xand I think about audio tools like Antares and Melodye for auto tune correction for singers that can’t sing. So with the continued development of facial recognition, sound libraries, AI for the Mac and the eventual appearance of an 8K iPhone.
So here’s how it works – you get your footage (hey, someone has to have an idea !) – and you import it into the footage (this will all be Bluetooth 10G speed of course). And you select one of the new Apple templates (sales video, music video, action video) and with the help of AI, facial recognition, and pre built music libraries – the show edits itself ! There is even improved voice recognition for the “editor” (apple has renamed this “image constructor”)
and you can say “insert a dissolve here” or “fade out to black” – and it does it. No key strokes. Find that image of Tiffany dancing the routine ? No problem – facial recognition is fully functional, and finding those shots of Tiffany dancing away is a verbal command. Who needs those pesky Keyflow/CatDV/Axle programs anyway ! Now, its not innovative cutting edge editing – but hey, it’s a sales video, it’s a local band or wedding band, it’s people filming the school football game (which airs on the local cable channel and the schools YouTube channel).So guess what – you are all out of a job (except the guys in major shows in LA). Good luck with your new careers.
Bob Zelin
Bob Zelin
Rescue 1, Inc.
bobzelin@icloud.com -
Really interesting article, thanks for posting.
It might be that the traditional assistant editor job might be at risk. If the NLE of the future can organize based on word, phrase, shot type (FCPX has the Find People option now), shot location (GPS), balance the audio, base level grade and other “pre-edit” tasks, is there a real need for an AE? Will this system then be able to string together basic scenes based on Scene #, take # and comments (maybe string together all the “circle takes”)? I could come in after a shoot, ingest the footage, let the machine do it’s thing and come back later and have all that done without another salary on the project.
Interesting.
Scott Witthaus
Visual Storyteller – FCPX, Premiere
https://vimeo.com/channels/1322525
Managing Partner, Low Country Creative LLC
Professor, VCU Brandcenter -
[Scott Witthaus] “I could come in after a shoot, ingest the footage, let the machine do it’s thing and come back later and have all that done without another salary on the project.”
Of course, in that scenario, you wouldn’t have the foggiest clue about your footage, how to actually structure the edit, or what gems were ignored by the AI. That’s because you’ve abdicated that role to the machine.
https://www.npr.org/2018/06/30/624373367/more-states-opting-to-robo-grade-student-essays-by-computer
– Oliver
Oliver Peters – oliverpeters.com
-
Meh.
Scenario: Among our editors: Jane likes things one way because of how she thinks. Robert likes things organized quite differently because of how he thinks.
They’re both great editors – but each resonates with a different process.
The producer (if they’re smart) goes with one or the other depending on how well they fit the projects goals and objectives – just like you pick any artist or designer based on their style.
The AE needs to set up things differently for each. THAT is where judgement beats machine learning – at least for the coming generation or two.
On the long run, bets are off. But I feel pretty safe for the foreseeable future. And I think AES should too.
Just focus on building your judgement chops – not so much the process ones.
FWIW.
Creator of XinTwo – https://www.xintwo.com
The shortest path to FCP X mastery. -
AI/Machine Learning always seems to be framed as a replacement for humans, when, in fact, it’s more like augmentation. And perfection always seems like an elusive 5 years out, which never seems to come.
When it comes to editors, we all look at assistant editors as vulnerable. Yet, how many here on this list – or in general – actually work with assistants? I rarely do. I’ve only done so on features and then not at a Hollywood level. There, assistants do quite a bit more than the grunt work that AI is supposed to replace.
Color me a skeptic, but here are some notable AI/Machine Learning failures – to date in our world:
1. Speech-to-text. A great tool that works well when the audio is clear and the speaker talks slowly with good enunciation. The minute you have 2 or 3 people talking over each other quickly, and with a poor recording, the quality of the transcription drops significantly.
2. Adobe uses machine learning to auto-duck music under speech. But, it’s very mechanical and cannot distinguish the differences in musical intensity. It’s a starting point, but only marginally IMHO.
3. If I have a silent video file with a slate/clapstick and a separate audio file with clap. No matching audio nor timecode. There is no current automatic way to sync these two up other than by manually setting in-points.
4. Shot detection can determine singles and two-shots, but it cannot correctly determine what the shot is when a shot starts on a single and then later zooms out to become a two-shot.
Maybe in 5 years ☺
– Oliver
Oliver Peters – oliverpeters.com
-
[Oliver Peters] “Yet, how many here on this list – or in general – actually work with assistants? “
I work with assistants and there are invaluable to me. As pointed out earlier, each editor will ask their assist unique tasks dependent on their organizational style. Also I rely on them for dialogue breakdowns and scene stringouts, and sometimes to do rough assemblies of scenes if Im working from dailies. Dont think a robot is anywhere near being able to replace them seeing as how specific some task are.
-
since this is an “OT” thread, I will express my opinion – that’s only an opinion.
What is sad about the possible reality of this is that when you “get into the business” – no one says “ok, you are an editor. You start off as an assistant. The main editor does the creative part, but the assistant (who eventually becomes a top assistant, who is very technical) does everything else – ingest, transcoding, delivery – everything but the creative process. And there are many guys (think Sam Mestman of Lumaforge) who were exactly this person – people like Sam Mestman know more than most about workflow, and all the technical aspects of editing.And the same applies to camera operators, and the camera assistants, who are the guys that go to the rental houses, get all the setups done, and make sure everything is working. What the assistant is not doing is the lighting direction and “pointing the camera” to do the creative work. But if he is a good assistant, he will eventually become the main camera operator.
Same with a top assistant editor. They become the editor. They make the connections.
What I see now (and as the assistant role disappears, as budgets disappear) is that there are fewer and fewer jobs for assistants. And so the “kid” that gets out of school – well, he owns a Mac Book Pro, and he knows how to use Premiere or FCP X (or both) and he cut a few “shows” for school, and his friends – he tries to start his own company, giving away his services. Same with cameras like the Red Raven. Kid can’t find an assistant camera gig (because there are few, and in the future there will be none according to this AI speculation) and so mom gets him the camera, and now he tries to start his own company, and gives away his services.
Everyone says “this levels the playing field” – but what I see is what I saw in the recording studio industry. THERE IS NO recording studio industry anymore. With rare exception, all the recording studios went out of business (LA post audio is the exception) and a small handful of studios with a reputation. And even those studios (like currently in Nashville) have to charge rates lower than ever.
So I believe that as AI starts to actually work, all that will happen is that it will make it much more difficult for young people to “get into” the professional market, and these young people (who are flooding these schools like UCLA and Full Sail, etc.) will get out, and they will all try to compete, and ultimately “give away” their services. You can say that “well – this is the inevitable change of the industry”- but all I see is the same demise for us that happened to the recording studio business, and has pretty much happened for a lot of the post production facilities in the United States. As AI develops, this demise will continue, and the industry as we know it will go away – except for high end jobs for major shows and big budget feature films.
When someone says “all you need to be successful is a paper and pencil and a good idea” – does that mean that it’s all going to come down to your 8K iPhone, your 2022 MacBook Pro, and a high speed internet connection ? Is that the future of post production ?
Bob Zelin
Bob Zelin
Rescue 1, Inc.
bobzelin@icloud.com -
[Bob Zelin] “What is sad about the possible reality of this is that when you “get into the business” – no one says “ok, you are an editor. You start off as an assistant. The main editor does the creative part, but the assistant (who eventually becomes a top assistant, who is very technical) does everything else – ingest, transcoding, delivery – everything but the creative process. “
But Bob, that’s a model that only exists in dramatic TV and film – mainly in LA/NY/London/etc and a few other places that work in Reality TV, not in major centers. For most editors who work in news, commercials, corporate – if they started in the tape days – the starting point was a tape operator. Those duties were a lot different than assistant editing. Since the nonlinear days, many of those jobs simply went away, so the foot-in-the-door job came somewhere else – production assistant, runner, operations, etc.
I think today, people entering the field come from a more rounded background, where the experience isn’t just editing. It’s a combo of production, graphics, and post. So that starting point is varied for the majority of the business.
[Bob Zelin] “does that mean that it’s all going to come down to your 8K iPhone, your 2022 MacBook Pro, and a high speed internet connection ? Is that the future of post production “
Unfortunately, for a lot of the industry, that answer will be ‘yes’.
– Oliver
Oliver Peters – oliverpeters.com
-
[Oliver Peters] “I think today, people entering the field come from a more rounded background, where the experience isn’t just editing. It’s a combo of production, graphics, and post. So that starting point is varied for the majority of the business.”
While it’s true that young people entering the business have a more varied skill set, the need for mentoring remains. Being able to tell a good story in an efficient manner does not come naturally to most. There may be young folks with great technical skills in multiple programs but the technical is always secondary to the storytelling.
Windows 10 Pro | i7-5820k CPU | 64 gigs RAM | NvidiaGeForceGTX970 | Blackmagic Decklink 4k Mini Monitor |
Adobe CC 2018 |Renders/cache: Samsung SSD 950 Pro x2 in Raid 0 | Media: Samsung SSD 960 PRO PCIe NVMe M.2 2280 | Media: OWC Thunderbay 4 x 2 Raid 0 mirrored with FreeFileSync -
That’s nothing – this baby felt like the money moment in terms of what’s coming down the pike.
Autonomous, completely accurate arrangement of the cast in order of appearance, rule of thirds applied, semantic shot analysis, detecting camera zooms and pans, detecting motion blur, on screen violence, depth of field, etc.
It’s for analysing stills, but it’s the best indication of where it might be going. The point about it is that Netflix has a strong economic interest in optimising and automating high quality show poster shots. A really strong interest given the content scale problem. They’ve done that, or rather, they’ve successfully adapted parts of the currently red hot domain specific machine vision learning.
It’s… difficult to think of a situation where they’d dump the necessary tens, if not hundreds of billions into generating a fully autonomous intelligence with the correct cultural nuances to replicate a warm body in a seat knocking out a thirty second promo tho – let alone edit together the collected rushes of a film project they’d already dropped millions into, when a reasonably trained human being can knock it out while making decent banter in the room with the director. And the bill is a lot lower.
If you read the Verge lately, in terms of autonomous cars, they’re coming as close as they can to calling absolute bullshit on the likelihood of unsupervised learning in unpredictable environments: that’s general AI, and solo cars need that. Sorting stills, that’s domain specific AI, basically a trained task nub – that’s happening.
But sublimating the gestalt of a trained human editor into code is wandering into general cognition AI – if we get to that, we’ve basically met/made/become god and editing jobs are the least of our worries.
Reply to this Discussion! Login or Sign Up