Can AI make a YouTube comedy news show?
I decided to do something crazy, to set myself a challenge - make a topical comedy news show for YouTube and TikTok, with help from AI.
Why?
Because being funny is difficult. So, if AI can do that, it can make ANY content, can’t it?
AI presenters, AI voices, and an AI script.
Did it work?
We’ll get there in a minute.
But first, let’s do a little “State Of The Nation”. Or a “State Of Video AI”.
Where are we at the moment?
I’ve started making videos for clients with the help of AI. How much AI we use depends on their requirements.
And that sums up where we’re at with AI video at the moment - all over the place.
Probably everybody who makes videos has used ChatGTP to outline and/or write a video script. And I have to say, as far as it goes, it’s amazing for that - at least for brainstorming ideas, outlines, and a first draft.
Other video tools are hit-and-miss.
There’s software where you enter a text prompt and the AI comes up with actual, original video footage.
Sounds amazing.
It isn’t. Not at the moment.
I’m, of course, thinking about RunwayML.
Unless you want to spend endless hours in post-production, fixing human heads that might start off looking realistic, but then morph into an egg (and I mean one from a hen with a shell) it’s just not great. Obviously that will change.
AI voices, on the other hand, have come a long way. They can now sound very realistic.
I say that, but I’ve spent hours looking for good ones. Even the suppliers of the good ones, have bad ones. And even if you find a good voice, it generates different results each time. Because it’s not like a recording. You can enter the same text twice and the voice will give differently intoned renditions.
So you do need some patience, and even be prepared to change words in a script.
I tried putting “Ah” into one of my scripts to make it sound conversational. But no matter what I tried, the AI voice sounded like she was at the dentist.
What about AI presenters, or “AI talking avatars”?
Again, we are so close. And again, it takes a lot of searching. But I found a couple of companies that are producing great results. The mouth syncing is almost perfect, although sometimes the lips don’t quite close as much as they should.
And they’ve really worked on the gaps between sentences. When an AI avatar has nothing to say, it needs to do those tiny head and eye movements we all make, in between sentences, that signal our sentience. And again, most of the time now, it’s great.
Back to the comedy news show.
When choosing a format, I tried to play to AI’s strengths.
I thought that if my AI presenters were playing news presenters, it would mask them having to “act” the script.
I based my idea on Chris Morris’ seminal comedy show “The Day Today”, taking real news stories but twisting them inside out for comic effect.
“News To Me” was born.
Now the script. Would AI be able to write the script?
First of all, I picked the news stories of the day.
Then I asked ChatGTP to make them funny, or words to that effect.
It tried. But there was something lacking. A lot lacking. Humour is subjective at the best of times, but its gags were largely just tired clichés.
For example, Ariana Grandé was divorcing her husband. And this is typical of what ChatGTP considers satire: “Ariana Grande's marriage to Dalton Gomez lasted just one year, which is longer than the average lifespan of a reality TV star and shorter than the lifespan of a TikTok trend.”
So I ditched ChatGTP completely and decided to write the script myself.
But I used AI presenters with AI voices.
I put out a series of YouTube Shorts for two weeks and they got over 50,000 views without any promotion or existing fan base. It was a brand new account. There were only a couple of comments asking if this was AI. The majority didn’t notice or didn’t care. One person was just entranced by the presenter: “I could listen to your voice all day”.
And now the main show is ready.
Watch the long version yourself, it’s 4 minutes long, and make up your own mind.
I’ve continued publishing shorts on the YouTube channel and on TikTok. Combined views in the first four weeks are 250,000.
The thing that stands out most for me is that I couldn’t have made this kind of show on my own without this new technology. Not having to worry about the process of filming meant I could focus my energy on what was truly important to this project, the scripts.
I massively enjoyed making this show and will probably keep making it, time permitting.
The truth is, I could not have made this without AI. The amount of time to set up the filming would have been prohibitive.
AI has bought me the time and creative freedom to do this.
When I started making it, I wasn’t sure whether I’d feel involved in the creative challenges, or feel like a passenger. I was very much not a passenger.
Will things change in AI content production? Of course. But before this experiment, I was worried about AI, and now I’m excited by it.
Subscribe to the THIS IS ME youtube channel to see the show’s progress and what else I do with it. Or go to the TIKTOK channel.
You can also sign up to my newsletter to keep up-to-date with adventures and tips in AI content.
Recommended Software
AI AVATAR
HEYGEN - this is the company I used for the “News To Me” presenters. The facial movements and lip-syncing are really good and getting better all the time. You can also use simple photos as the basis for an avatar, though this isn’t quite as good.
SYNTHESIA - This was the other AI avatar company with animations that impressed me.
AI VOICES
ELEVENLABS - Most people recommend ElevenLabs for AI voices. And this is what I used for “News To Me”. Beware though, you often have to do a lot of “regenerating” to make the dialogue work. And you pay every time for this. Their voice cloning, however, is excellent.
PLAY-HT - This was the only other company with AI voices that impressed me. And they have the advantage over ElevenLabs of not charging you for every regeneration of the dialogue. You only pay when you’re happy with the final dialogue and download it.
OTHER SOFTWARE
PICTORY - This is an AI-powered online video editor. If you’re looking to make simple voice, text, and stock footage videos this makes it surprisingly easy and fast. Now, don’t get me wrong, Pictory is no match for Premiere, Final Cut Pro, or Da Vinci Resolve as a video editor.
But it does one thing well.
It matches each of the lines in your script with stock footage from Storyblocks and Getty Images. That’s a lot of footage.
All you have to do is edit the transcript of your voice-over by putting a new line in where you would like a new shot. The software then matches it with a shot from this vast library.
If you don’t like the shot, you just tap on it, then fill in a term on the search bar, and it’ll give you a selection of new shots. Really simple, really fast.
Now, very often I didn’t like the A.I.’s selection of shot. In fact, I probably only used around 20% of its choices (mainly for aesthetic reasons). But it was so easy to swap in new footage.
The problem with making videos from stock footage is that all the shots are lit differently and use different colours, so the video can look less than cohesive. So I took the finished video from Pictory, did some quick colour grading in Da Vinci, slapped on a LUT, and I was done.
Now, I know the online image editor Canva can make 100 videos in no time at all with its bulk creator, but the quality of the ones I’ve seen has been variable.