Dumme, a startup putting AI into practice in video editing, is already generating demand before going public. The Y Combinator-backed company has hundreds of video creators testing its product, which leverages AI to create short form videos from YouTube content, and a waiting list of more than 20,000 pre-launch, says he. Using a combination of proprietary and existing AI models, Dumme’s promise is that she can not only save editing time, but also – and here’s her big claim – do a better job than manpower. contracted (human) labor that is often tasked with more menial tasks. video editing work, such as shortening long content for publication on short platforms like YouTube Shorts, TikTok or Instagram Reels.
Founded in January 2022 and a participant in startup accelerator Y Combinator’s winter 2022 program, Dumme co-founder and CEO Merwane Drai said he was initially focused on building a search for video. But about six months ago, the team realized that a better product might be to reuse the same AI models they were developing to edit video clips instead.
Joined by co-founders Will Dahlstrom (CPO) and Jordan Brannan (CTO), all with AI backgrounds, Drai realized that Dumme may have landed on the good market after their app went viral, making crash their servers.
“We weren’t really expecting it to get much traction or anything, so we just put something there,” says Drai. “Then what happened was that overnight we woke up to overloaded servers – like nothing was actually working. So we wrote everything down and created a kind of waiting list” “The next morning we probably woke up to 5,000 people, which was interesting.”
The team later discovered that a TikTok creator posted a short video about the product, which sent a flood of traffic to his site.
“It actually never calmed down,” Drai notes.
The product, pronounced “dummy”, appealed to creators because it aimed to simplify and speed up the work of video editing.
Using Dumme is as easy as the name suggests. To get started, the user pastes the link of a YouTube video, then clicks “generate” and the AI will produce a number of short videos showcasing highlights of that ingested content. The company says it uses YouTube as a source, instead of supporting raw video footage, in order to outsource content moderation – i.e. if it’s allowed on YouTube, it’s allowed in Dumme.
The processing time and number of resulting clips will depend on the length of the original video.
But as an example, an hour-long video podcast might take about 20 minutes to process, and you’ll start receiving clips after about five minutes, Drai says. Once complete, creators can download the music videos, which are less than 60 seconds long by default, and upload them to any platform that supports short-form content, such as YouTube Shorts, but also other platforms. forms, like Reels or TikTok.
How all of this works on the backend is of course much more complex. The company says that, initially, Dumme will learn as much as possible from the source video through metadata. He then transcribes the video and tries to understand the semantics of what is being said while watching the images to try to decode the emotions of the person speaking. These results are correlated and passed to a language model which tries to determine which parts of the video are worth cutting. This is then transferred to another model that tracks active speakers and handles cropping.
Dumme says it works with existing AI models like GPT-4, a fine-tuned version of Whisper, and others he’s built in-house, like the model that tracks active speakers in a video frame. . One of his models is also trained on a bunch of YouTube shorts to know what makes a good opening hook to attract viewers. And, although not yet live, the team is also experimenting with an open-source model, LaViLa from Facebook Research, to better understand the context of the video.
The AI work is done on cloud GPU provider CoreWeave, not on AWS because it’s more affordable, the company tells us.
Because Dumme relies on AI that processes spoken words, the technology isn’t appropriate for things like long gameplay videos or others where people aren’t speaking. Drai says the startup is initially targeting YouTube creators, podcasters and agencies — the latter they believe would be the best bet to monetize the product.
Agencies today, Drai says, often outsource this type of work with hit-or-miss results.
“They just pay contractors in cheap jurisdictions to edit their own content. And the problem is that it’s actually still quite expensive and it takes a long time — it takes weeks, not minutes,” he says.
When asked what he thought about creating technology that would put people out of work, Drai wasn’t worried.
“The way I think about it is that ultimately…I think it’s like telling me that the math teachers are going to [be put] out of work because there is something called a calculator…”, he explains. “People will adapt. And then there will be someone who will teach you about the calculator, right? So I think it’s just a matter of adapting to that,” says Drai.
Currently, the pricing being considered involves tiers where a business would pay $0.40 per minute of processed video, while smaller creators can instead opt for a monthly subscription capped at 10 hours of content per month. (These numbers are subject to change.) During testing, the product was free to use.
Early adopters used Dumme for a variety of edits, including generating clips from their video podcasts to post to Shorts, as well as trimming other new videos and playing their past catalogs.
The product appears to be competitive with other AI technologies on the market, including that of creator company Jellysmack, which leveraged AI to turn longer YouTube videos into shorter ones, cutting them, by resizing and optimizing them for specific platforms – a result of its acquisition of Kamua in 2021. Other tools doing similar work include things like Opus Clip, Vidyo.ai, Detail, TubeBuddy, Wisecut and others. The extent to which Dumme succeeds or fails will be to outperform its competitors in terms of quality of work and cost – parameters that have yet to be determined.
But some investors are betting on Dumme. Prior to launch, the startup raised a $3.4 million funding round from Y Combinator, Caffeinated Capital, Max and Nellie Levchin (via SciFi VC), Suhail Doshi, Nico Chinot, Protocol Labs, Chris Puscasiu and d other angels.
Given the interest and the large waiting list, Dumme says he aims to board around 500 people each week. TechCrunch readers can skip the line using invite code TECHCRUNCH while slots last.