YouTube's next AI Goldrush?
On our cast 👆 we give out digital business ideas every week, but this week we think we got a concept that is uniquely interesting, we call it "Programmatic Videos".
In this video, Tom and I discuss 2.5 ideas you can take and start making money. We hope you like it 😉
Trending on Indie Hackers
🔥 Roast My Landing Page
IMO we need to wait a bit to have a good-quality AI for videos, anyway, I think in the following years we will see a massive increase in top-ten videos and Cartoons/After Effects ads done by AI.
As we show in the video, there are already some creators doing this.
If you use tools like https://www.remotion.dev/ I think you can create decent videos right now.
Yes, I know. But this product is not for the masses, and the quality is decent as you said. Is quite hard to imagine this trend at the same level as the text-to-image or GPT trend...
Let's wait and see :) The question is who will be the winner at the end.
Thanks for the talk, that's exactly I was thinking and trying to develop an AI tool which generates the videos from the text using generative ai and script (TextToVideo Bot)
Thank you for the reply. How is TextToVideo Bot going?
ChatGPT is pretty good at scripts, but I haven't seen any tools that can generate usable video content from a script and didn't see any mentioned when I skimmed your video. I recently explored this a little myself, and the best reviewed tools in that department were still extremely lacking.
In the first example you showed I suspect that guy has experience writing video editing software and put significant work into a custom script that pulls from a list of coding Q&A and can generate only the one specific format it's designed to spit out.
The first example can be achieved with very little programming knowledge (using no code scraping tools then using something like VAs/Fiverr editors to stitch them together - or using basic ffmpeg commands).
Regarding libraries there is moviepy and remotion that come into mind.
Both are nice, but I agree that we will probably see better advancements soon. Not to mention that Generative videos might be coming soon too.
Better libraries = Better SaaS solutions.
He would be millions of dollars in the hole if he paid people to edit the videos so I don't think he's doing that. There's something off about the "human" video, I think he's recombining a handful of clips somehow or maybe using something like synthesia.io.
Basic ffmpeg commands are enough for stitching, but to actually create the text scrolling etc. will be quite a bit of scripting work as I said. The other libraries you mentioned still need a lot of scripting work to produce content, so they are in no way competitors to Pictory etc.
So my conclusion remains the same. In order to do something like this you need a lot of scripting work. You'll also have very little flexibility, and be extremely limited by lack of real footage, meaning the videos will be less engaging. Notice he only has one video (out of 2M!) that would have made him more than a few dollars.
Did you try using ChatGPT to code some animations and examples using moviepy/remotion? (I did and it did a decent job) - granted, you still need to know some coding to get it working completely, but still it gave an amazing result for me.
The aggregate of videos make money, no?
Yes, I agree that if you put in the work writing scripts/etc this can be viable.
I do think good stock footage and/or human faces are needed to make engaging video most of the time though. The only animated videos I've seen do well done by people who specialize in animation and have built up the entire skillset needed to do it well.
Before utilizing ChatGPT, I began creating automated YouTube shorts without relying on any AI technology. However, after incorporating ChatGPT, the process of obtaining data and scripts for the videos has become much more convenient and efficient.
That's awesome!
Do you have an example to share with us? (no worries if it's under the radar).
Just an experiments with news feeds, trying to implements in other dynamic categories youtube@news-13.
any good youtube niche you can suggest??
We are going to record at least 2 more videos with programmatic videos as the underlying business in the coming weeks/months.
A few things that come to mind are AI mock interviewer and Gaming (+streaming) highlights.
I am not sure how long this gold rush would last. There is always going to be value in connecting with your audience with your face
I think as long as we advance forward tech-wise with things like https://www.synthesia.io/ (that will get better over time) we will still have these automated outputs.
I agree that a personal touch is relevant, that is why Tom and I started Up for Grabs and we show our (real) faces :)
I can't imagine Youtube tolerating auto-generated spam AI videos (like Roel's) for very long. I know other channels with 1M+ spam videos have been shutdown.
But how is he getting away with it? My only guess is the in-person intro segment - I don't think these are AI generated, more like he recorded 20 intros and just reuses them, and somehow this is enough to fool Youtube's spam filters.
Does anybody know the click-through-rate for affiliate links in video descriptions on Youtube? Some quick googling suggests 1-3% but I can't find any in-depth analysis with evidence.
Also @agam install the uBlock Origin extension you'll never see a Youtube ad again.
I completely agree! But, it does beg the question, why doesn't YouTube write anything about it in the Terms? Why aren't they shutting Roel down already? It's pretty well known (https://youtube.fandom.com/wiki/Roel_Van_de_Paar). And the heuristic can be as simple as it isn't reasonable to upload that many videos that quickly - so there might be something there for YouTube.
My personal opinion is that you should provide value, just like in Programmatic SEO (Google will and does crush a lot of spam websites).
I'd also love to hear about the CTR!
P.S. I have uBlock origin! I just recorded on a different browser so my history and recommendations won't show up as they aren't relevant and I don't want them out there. I'll install it for next time though!
I would argue it is too early for videos. I would personally turn off immediately if realized it's an AI video.
The personal aspect is why most people watch YouTube over TV. With AI, we back to large TV generic approach.
But some day, yeah, it probably will be the case.
What if you don't realize it is AI generated? The whole point is that you can use it as a marketing or revenue channel - assuming you found a way to make valuable content that looks "good enough".
Yeah, but such videos don't exist yet. There is no good YouTubers that are actually AI.
Even pictures are easily spotted. In 5–10 years? Maybe, but not now.
My estimation is you'll start seeing decent videos this year. It might just be superb animations that are generated, not necessarily AI models.
I am sure I won't be watching those though. It won't be a decent quality/value/personal.
Let's just wait and see. But judging by the history, people, especially who exited about that, always overestimate these things. I would gladly to be wrong.
You still often can tell when the text is written fully by AI. I would be curious to see at least text-based content with a good quality.
Let's wait and see then 🧙♂️
Kinda reminds me of Netflix's programmatic thumbnails but on steroids haha
A reference of that story for people that don't know:
https://recosenselabs.com/blog/how-netflix-uses-ai-and-ml-to-determine-which-thumbnails-they-use
Yeah, it is similar but not the same, it's more similar to Programmatic SEO (https://unzip.dev/0x003-programmatic-seo/)
So in Netflix's thumbnail case they generate it to get more clicks on the shows. In our case we are trying to target textual keywords. But! you can achieve the same thing with midjourney auto generating your thumbnails for similar videos and seeing what works better in theory.
YouTube has been investing in AI technology for a variety of purposes, including content recommendation, video analysis, and moderation. AI advancements have the potential to significantly impact various aspects of YouTube, such as video recommendations, content filtering, and user experience.
Yes indeed, but here we talk how you and me can leverage ai to create content at scale on YouTube. What are your thoughts on that?