A.I. is already widely used, but how can we use it better?

[Steven Makari]

Hello,

You’re absolutely right; A.I., or Artificial Intelligence, has become an integral part of many industries, including the creative sector. It’s often working behind the scenes to enhance various aspects of content creation and distribution. Your experience with transcribing and translating a 9-minute interview highlights how A.I. can streamline time-consuming tasks.

A.I. tools like those from Nova and big players like IBM Watson and Adobe have made significant strides in automating and simplifying these jobs. However, you’ve correctly pointed out that there can be challenges with context, especially when it comes to nuanced or creative content.

The role of A.I. in the creative process is indeed a topic worth discussing. It can be a valuable tool in terms of efficiency, but there’s also concern about its potential to replace or homogenize human creativity. Striking the right balance between using A.I. as a creative aid and preserving the uniqueness of human expression is a crucial conversation.

In the end, there’s no definitive right or wrong answer. What we can learn from A.I. is how to harness its capabilities while maintaining the essence of human creativity. It’s an ongoing journey, and the creative community will continue to adapt and evolve with these technological advancements.

[Finney Barndlow]

Totally agree, A.I. is like a tech wizard behind the scenes, making things happen faster. Definitely a lot of potential there.

Imagine A.I. jazzing up brainstorming sessions, helping with music composition, or even suggesting color schemes for that next masterpiece! It’s like having a creative sidekick. Sure, it might stumble, but it’s a learning process for us all. At the end of the day, half of the job is to instruct your AI tools properly and tweak them until you get the best result.

[Mads Nybo jørgensen]

Interestingly to observe how many companies that has migrated from “Machine Learning”, to overnight call it “A.I.”.

Since A.I. became a thing, there is a new twist involving:

Creation of deep-fake photos and videos with just a few lines of text.

Limited data-sets as A.I. can only generate from what it can see and learn, mostly from the internet.

No consideration to all of the creators whose original work are being blown to pieces and put into new works (although some might argue that the music industry has being doing this “manually” for decades. Nevertheless, what is an original piece of work, and what do you define one that has been made up from over a 1,000 pieces of work?

I ask myself whether I should encourage my kids to learn Python so they can programme A.I., or just let the A.I. program itself?

Two things that we can be sure off is that A.I. is here to stay, and the more abuse that it used for making, the bigger chance is that our law-makers will put heavy regulation in place for using and identifying A.I..

Atb
Mads

[Kendall Hoover]

You are correct. AI will help our jobs out here in the media creation space and we will be able to make better content faster because of it.

[Robbie Janney]

I believe AI is going to assist creatives, and unfortunately for many MBAs that are looking to “cut costs wherever possible (usually in the creative department)”, AI will not be replacing people anytime soon.

How do I know? Well, one, i’ve been an editor for 10 years, and two, I work for a startup who’s primary goal is to create AI tools for video editors.

I’m headfirst in the AI video editing sphere every weekday, and most every weekend.

So it surprises me when some video editing AI companies claim that “you won’t need a video editor anymore” – most of the final products that they produce look like every other one that comes out – it’s a script that is followed and doesn’t have any of the real branding or visual look of videos edited by pros.

Like, i’ve seen the current limits of AI – I talk to our AI scientists every day about the subject.

We’re very, very far from a “God Button” of any sort that can spit out an edit exactly how you want it.

BUT, AI has advanced enough that it’s super helpful in some cases.

For example, we’ve got an automatic social media highlight selector (Reels Finder) that finds the best 1 minute sections of your long form video. You can prompt the AI with a specific topic, and it’ll find the best reels according to your instructions.

The cool part is that you can edit the reels with transcript editing, add captions, and then export the reels to Resolve/FCP/Premiere for polishing and mo-graph support.

We also have a soundbite selector, which grabs all of the best takes from raw interview footage that you can export as a timeline to Resolve/FCP/Premiere.

We’re also working on some cool stuff like auto-logging and bin creation using computer vision, and rough cut creation using prompts.

It’s all in service of making life faster and easier for editors, since at the end of the day, the human decisions that go into the final cut is what makes a video stand out.

We’re free at the moment, so try it out if any of this sounds interesting. (Storylines.video)

 

[Mads Nybo jørgensen]

Hey Robbie,

You make some good points.

But your company is also promoting this:
“Trained On Your Edits

You allow Storylines AI to learn your style of videos, tone, & voice, the foundation to build a bespoke video editing AI, just for you.”

Not neccesearily directed at your firm, but at what point will an untrained editor be able to use my unique style of editing, without me getting a cut, or even knowing that an A.I. has analysed how I edit, and sell that as their own?

The same for Motion Graphics, After Effects, and animation.

Adobe just recently tried to force through T&Cs allowing them to do just that, take ownership of my creative work.

Personally I am not afraid of this, but I am also determined to hold A.I. companies accountable to track when they use my work, and to pay me for it.

Yes, in my juristiction you can not copyright creativity, but you can ask for a credit (and money) where you work or work practices has been used. As the A.I. collects all the data, it also would be able to create a “list of contributors”, right?

Just encouraging discussion, before the politicians decides to legislate.

Atb
Mads

[Robbie Janney]

Hey Mads!

Excellent questions, thanks for the response.

So, when Storylines says “Trained on your edits”, that means that the AI that we can create for certain workflows is trained on a specific person’s workflow – but that information is not used to feed the general AI.

Think of it like this: we’ve got general AI tools that are available to the public on our site. That AI was trained by us on our edits and they complete a certain type of task that can be widely used by just about anyone – social media highlights and soundbites, specifically.

BUT, say there’s a company that has a certain type of video that it creates in bulk. For example, product information videos that they use for their website. We can train an AI on that specific problem – (the bin structure the editors use, a basic rough cut creator, lower third generator according to the transcript, etc.) all curated for those specific tasks that the editing team wants to speed up. So, a separate pipeline is created that is trained for that specific system, and not the general AI.

Since video creation varies so widely and the general knowledge that there isn’t a “edit my video” button that works for everyone, we have a general AI that we train on our own information and bespoke AI that uses the general model, but is modded for specific workflows of editors/companies that have asked us to train on their videos (for their use only, of course).

I think that’s the best use of AI – those specific pockets of tedious tasks that come with an editors workflow that can be offloaded to an AI, trained on that specific editor’s instructions so they can get to the more creative, intuitive stuff faster.

[Mads Nybo jørgensen]

Hey Robbie,

Sorry, July has been crazy for me and suddenly nearly 3-weeks goes by before I empty my in-box, and there your reply was “hidden”.

I like the idea of training the A.I. to make workflows more efficient.
As I am sure that you know, when finishing an edit, that is where the real hard work starts on Compliance, Q.C. and various deliverables depending on platform and client, followed by another round of Q.C..

I wonder whether a simple macro like in Excel (called “Action” in Photoshop) could do the job?

It will be interesting to follow what you do at Storylines.
Challenges, as your bigger competitors have found: The cost of Cloud Storage, analysing footage accurately, then edit it, and export to something meaningful to a video file and/or edl/xml for craft edit NLE.
Before you calculate the power usage, which currently is the elephant in the room for A.I.

I would suggest that Storylines sooner, rather than later, put a cost (pricing) on the website for this. Not least as your average “Pro” project rushes can easily range in size from 10GB up to 20TB+.
20TB on Google Drive costs near to 80 GBP a month + then you’ll have processing + the edit.

Granted, that is not likely the market that you are gunning for as a start-up for professional editors.

But as you are mentioning Auto-Multicam and Sports highlights on the site, why not create a plug-in that can work with Broadcast servers and inside the current editing tools?

We have just had the UEFA European Football Championship (won by Spain). The Olympics is starting in Paris next week. For each major event, there is an entire team with their own “truck” dedicated to making highlights and replays whilst the event is going on.

Same goes with conferences, music and other events.

In a short sentence, the business could get killed by cost of storage and uploading before you’ve even started the process of A.I. editing and Post-Production. Unless you sit on top of that, inside the production as it happens, using their servers.

Just a point of view.

But I’ll be eager to take your first full release for a spin.

Atb
Mads