AI for video editing - why is this use case being overlooked?

Jon Reed Profile picture for user jreed March 25, 2022
Summary:
Why do vendors overreach, while viable AI use cases get overlooked? AI for video editing is one glaring example - though with options like Pictory.ai and Riverside.fm, that's starting to change. Here's what I tested for my video show.

jons-video-show
(a screen shot from my first "AI-edited" video)

I know what you're thinking - pigs fly again. Jon is pitching a hypothetical AI use case. Yes, I have scorched AI advocates (and marketing vendors) for overreaching on what is possible with "AI for content creation." (Can AI displace content creators? For B2B content, the answer is no - but with a disconcerting asterisk).

Just this week, I tested another AI tool for "writing engaging marketing copy" that generated exactly what I expected: tone deaf verbiage. All the right keywords, none of the impact.

Here's what baffles me: why are we pushing AI for generating creative copy that it (mostly) stinks at, when we could be pursuing content-related use cases that AI is much better suited for?

AI can be useful, I think, for proposing marketing taglines, keywords, and topic ideas for consideration. But writing actual copy? That's a great example of an overreach. Just about every week, I get a PR pitch for an elaborate AI-for-content platform. But it's rare to get a pitch about a content problem AI might actually solve.

B2B video editing scenarios - could AI help?

Here's one: AI for video editing. Even for content teams with deep pockets, video editing remains a cumbersome, expensive chore. For independent video producers and "influencers," it's even more time-consuming.

Scenarios:

  • You produce an engaging webinar series with industry experts. Each webinar is an hour long. You get some replay action, but how many folks have time to watch the whole thing?
  • You have a live video show series. The shows run long. You want to share highlights from the replay on social media. Have fun producing those clips on your own time.
  • Your virtual event generated some really good replay content. But how do you share video moments from that event on social channels, to lure replay sign-ups?

Longer-form video can be an engaging format, but replays hit diminishing returns. Yes, some diehards will watch the replay. But most of us won't be tethered to our phones or computers for replays of long-form B2B video content. Not when there are baby pictures and cat videos to see, "Billions" episodes to catch up on, and audio-only content that frees us to exercise, drive, and be a productive member of society.

To a modest extent, this problem can be solved by audio-only channels. My Busting the Omnichannel podcast series is mostly audio editions of video shows. But this doesn't solve for all - especially for events where visuals (e.g. a slide or two) played a key role. And it doesn't create shorter, engaging segments.

My favorite B2B video content creator, Brent Leary, is skilled at creating shorter highlight clips from his (excellent) shows - but even Leary has kicked tires on AI video editing options (Leary appeared on my reboot of my video show last fall, discussing the art of enterprise video). The AI-for-video idea is pretty simple:

  • Couldn't "AI" grab snippets of video shows for clips, perhaps aided by clues such as voice volume, laughter, etc?
  • Couldn't AI package those clips for social media distribution?

This is especially crucial for Twitter, where you can't upload a video clip longer than 2 minutes (unless you stream a show live on Twitter, in which case you can have a longer replay by default, but you can't re-upload that).

I definitely have this problem with my live enterprise video show - it's set up as a freewheeling live format, and can go a bit longer than an hour. The audience members are vocal; unlike most formats, they are a jugular part of the show. Therefore, commenting and watching after the fact is not nearly as fun. It would be ideal to share a couple of high points from each show. I probably have enough video editing skills to do it, but I definitely don't have enough time. Can AI save my bacon?

Riverside.fm's AI for video solution - "Clips"

With input from Leary, I went on a search. I was first tempted by Riverside.fm. Riverside is a video streaming platform that took a shot at the video editing problem, via their December 2021 offering of "Clips."  Why Clips? From the press release:

In a world where it's easier than ever to create content but harder than before to grab attention, people don't commit to watching a new show or series before seeing a part of it. So why not entice them with a short, exciting teaser?

Okay, we're on the right track here. Though I believe in providing more than just a content tease, but a satisfying segment in its own right. More on Clips:

It's a new way to repurpose your long-form content into shorter content pieces, shareable across social media to reach more people and entice them to subscribe to your channel. Instead of spending hours downloading, previewing, and trimming recordings in post-production, the new Riverside Clips feature automatically creates quick short content for all video creators. This way, creators can swiftly share teasers on various platforms to encourage more people to commit to their brand and longer-form content.

Yep, that's tempting. Riverside.fm adds:

Think of it, Hollywood already does this. No movie or big production comes out without a trailer to pull you in. Of course, filmmakers have a whole production crew behind them, but with Riverside.fm's, there's no need.

Now that's an overreach. I don't think AI would be good at scouring a one hour show and creating a highlight real from that show. At this juncture, that's too much to ask. Movie trailers aren't a "clip" per se. Trailers are a compilation of the best jokes/moments/action from a movie, compiled with tremendous care and production values (sometimes the trailers are better than the movies!). Trailers spare no expense. At this juncture, only a human can pull that off. But: you don't need a bona fide highlight reel of a whole show to share a good clip.

To the best of my knowledge, that's what Riverside.fm has done. It's not a highlight reel compilation service, but a clips service - thus its name. When you stream on Riverside.fm, Riverside automatically creates a collection of clips from the longer show. Just pick the ones you want, and share them on social channels.

If I don't have that completely right, don't blame me - I emailed Riverside several times for clarification, and didn't hear back from them. As such, I could not test drive or recommend their service. The pricing seems like it would encompass the needs of both enterprises and individuals, but I can't say for sure without more engagement from them.

Another downside to Riverside: at this point, you have to stream on their platform to use Clips (as far as I can tell). In my case, I stream via Streamyard to five locations, including my main live location, LinkedIn, which Riverside.fm doesn't serve. I also trust Streamyard's stability, and love the integrated commenting from all platforms, so I'm not interested in moving off of Streamyard. Riverside.fm could have a good opportunity to provide Clips as a separate service to their streaming, but that is not the current model.

Pictory.ai - an ambitious AI video editing solution

After a series of demo fails and false starts, I found my way to Pictory.ai. Pictory has several use cases: "short videos from webinars and text in minutes." For my needs, the webinar part obviously jumped out:

  • Create shareable highlight reels of your webinar in minutes.
  • Use highlights to promote your webinar on social channels.
  • Short video clips get higher engagement.
  • Harness AI to give your high-value content the high reach it deserves. No need to learn complex video editing tools.

Pictory.ai also has accessible pricing with a range of tiers. In this case, I signed up - no streaming on their platform required. For the "webinar highlight" AI editing service, you can upload your own videos, and  Pictory.ai does its thing.

The lowest pricing level isn't ideal. The next tier up provides the ability to put your logo on the clip, and also store an intro and outro, which is automatically affixed to each video. So what is the role of "AI" here? Via Pictory:

  • As your video is uploaded, Pictory’s advanced AI engine transcribes the video.
  • Our AI examines your transcript for its most consequential soundbites and highlights those sections of your webinar.
  • Easily confirm or change the automatically identified highlights.
  • Add your logo. Customize the colors and fonts of your captions. Add your own intro and outro.

In terms of "AI," a machine-generated transcript is not unique. At best, it's a low tier of AI. Pictory.ai's transcripts seem pretty solid, as far as machine transcripts go, but I haven't compared them closely to Otter.ai, which is my go-to for machine transcripts. But Pictory's next part, "Our AI examines your transcript for its most consequential soundbites" - that's very intriguing, and definitely advanced.

Pictory.ai is taking on a monster AI challenge. They are trying to create a true highlight reel of your video. I have not tested this fully. In my early Pictory.ai videos, I was on my learning curve, working through a few tech issues. Pictory's tech support was excellent; their user documentation is above average as well.

One really nice thing about Pictory: you don't have to put full faith in their AI. You can also edit the video based on the transcript. You can reject the AI's suggestions, or create your own. My early observations: the AI does a good job of identifying some high points, but they don't always stitch together seamlessly. That's to be expected. But to review the potential high points is very useful. Pictory's AI ambitions may be high, but they also allow you to modify what the "AI" thinks is important. So, in each video, you can decide how much you want to trust the AI.

As a consequence, you may have a bit of cleaning up to do around the AI selections in the transcript. So it's not fully automated - but I'm glad for that. It gives flexibility for when the AI might come up short, or when you have a different agenda. Yes, it's not seamless AI editing, but for now, I like that. And: editing videos via the text transcript is nifty - and fast.

For my early tests, instead of going with the AI highlight clips, I selected the video section that worked best as a standalone segment. Another small downside to Pictory: you do wait for Pictory to upload the video, process the transcript, create the "scenes," and then, later, export that video. The interface is also a bit tricky. I found that the on-screen transcript wasn't 100 percent accurate, not enough to simply use verbatim.

The nice thing is: you can suppress the subtitles if you want. I chose to suppress them on my first two videos, but only for expediency. Obviously, in general, subtitles are important for accessibility - though some platforms do provide them after you upload.

It takes a while to learn Pictory, work on your intro and outro, and get those uploaded. In my case, I used Canva to create my intro and outro clips. The first Pictory.ai video I produced probably took me 90 minutes, not including the intro and outro creation. Using the transcript editor, I selected a clip from the end of my conversation on the problematic future of retail, with Guy Courtin:

Now, this clip is eight minutes, which is hardly a teaser type of clip. But when you consider that the full retail tech discussion ran longer than an hour, pulling a nine minute video is much more realistic for replay. The replay could not be uploaded to Twitter, but it did get traction on LinkedIn. I had a small issue with the Pictory intro clip integration, but I solved that for my next highlight. The next highlight clip was via my show on busting HR project myths with Stacey Harris. Again, I picked the culmination of our discussion: "Does HR have a seat at the boardroom table? Why or why not?"

Subtracting Pictory's video upload and processing times, this one probably took me fifteen minutes to produce. To put out a clip with so little editing effort is big for me. Same story as last time: nice pickup on LinkedIn, but too long a video for Twitter.

After these two, I had to take a break from my video editing forays, which are limited to weekends. But next time, I will test Pictory's AI video clip, produce that one first, and see if it's worth sharing. I'll try to get under the 2:00 limit if I can, so that it is eligible for Twitter also.

Then I will potentially produce a longer AI version, or do another longer clip that is primarily one segment. Obviously, I could edit a few segments together, but there reaches a point where I'm venturing into time I don't have. Depending on my results, I may update this post, or, perhaps, do this for a new post.

My take

My intent was not to exclude interesting AI video providers. I did scour the Internet before writing this, however:

  • I was thorough, but not comprehensive. I relied on Google search, so to some extent, this was a test of search relevance.
  • I was focused on solutions that would be affordable for a content producer like myself. I did not look hard at enterprise-only solutions. What's an enterprise-only solution? If you don't have public pricing, but ask for a salesperson to contact you for a quote, that's enterprise-grade, and beyond my scope here.
  • This field is changing quickly - I could easily have missed something good. That's what the comments section is for. I remain open to testing new tools.

As I said in the disclosure below, I encourage readers to cast a wide net when evaluating tools, and never take my experiences as endorsements. I am encouraged to find a couple viable solutions for this need. I remain surprised how hard it is to find good, mature options for this, given the painful/expensive nature of video editing, and the obvious role AI could play reducing that pain.

Designing these solutions requires some humility in what AI is currently capable of, and a flexibility that allows creators to step in, if the process is too automated. AI will play an important role in content marketing in the next few years. However, I believe the majority of that heavy lifting will be in tools that do what machines are good at (flagging patterns, automating processes). If we can take a step back from the  fantasy that humans won't be able to tell what marketing copy is written by machines, then I believe AI can have immediate impact. I look forward to hearing your views (and experiences) on this one.

Bonus: the art of enterprise video show replay:

Disclosure - I am a paid customer of Pictory.ai. However, they upgraded my paid membership to I could access and test some premium features. I did not include screen shots of these services as I didn't want this post to come off like an advertisement. I encourage all who are considering these options to cast a wide net, there are plenty of approaches out there and I'm sure I missed a few good ones.

Read more on:
Loading
A grey colored placeholder image