I quit my customer service job to make AI videos full-time on YouTube. People don’t realize how expensive they are to produce.

This as-told-to essay is based on a conversation with Jonathan Laramy, 32, creator of the viral YouTube channel Chloe VS History, which features an AI-generated character named Chloe who travels through history.

The conversation has been edited for length and clarity.

A lot of people assume AI videos are created by typing a few prompts, pressing some buttons, and uploading the result.

I understand why they think that, but the reality is far more complicated.

A few years ago, I was working in customer service and spending my spare time experimenting with AI.

I had no formal filmmaking background, production training, or experience creating videos professionally.

In June 2025, I quit my customer service job to pursue my interest in AI.

I already had successful social media channels about history and was excited about AI and the prospect of trying something new with my YouTube AI history channel, Chloe VS History.

The channel is now earning me way more than my original customer service wage, but it took time to get there.

Chloe VS History follows an AI-generated influencer named Chloe as she travels to moments in history, like the Titanic’s maiden voyage and Pompeii on the day of Mount Vesuvius’ eruption.

Presented as a modern-day vlogger, she takes viewers inside historical settings and events.

My workflow combines several AI tools.

I use Claude for ideation and general script structure, generate images through PAI 2 (an AI software by Utopai Studios) using Nano Banana Pro and ChatGPT 2.0, turn those images into video clips with Seedance 2.0, and use an AI voice model to keep Chloe’s voice consistent.

Creating Chloe was one of the quicker parts of the process.

I generated between five and 10 versions of the character before choosing the one that felt most relatable.

I spent much more time refining her personality and dialogue so she would sound like a real influencer rather than an AI-generated character, and that attention to detail paid off.

One of the most common questions I get is whether Chloe is real.

While the videos on my channel may be generated with AI, they still require creative direction.

I don’t simply accept whatever the software produces.

Every scene needs to be reviewed, adjusted, regenerated, and edited until it works, just like a real movie.

Historical content, especially, creates additional challenges because modern AI models often introduce anachronistic details, like street lamps and sunglasses in my ancient Rome video.

That’s why a single long-form video can take weeks to produce.

The process involves scripting, image generation, video generation, voice work, multiple revisions, and post-production.

A long-form video for my channel typically costs between £300 and £800 ($400 to $1,070) to produce.

The final cost depends on how difficult the topic is and how many times scenes need to be rerun before they look right.

Unlike traditional YouTube videos, where a creator might spend money on equipment or travel, a large portion of my costs comes from generating and regenerating content through AI models.

Every revision has a cost attached to it.

It works out to about $3 per 15-second clip, but a single video can take 10-15 revisions, so the costs add up.

And those expenses are there whether the video succeeds or not.

Early on, Instagram wasn’t generating much income for me because the platform doesn’t pay for views.

TikTok also wasn’t a major source of revenue because I wasn’t accepted into its Creator Rewards program.

Even on YouTube, monetization took time.

So for a long time, I wasn’t making much money from the project.

It wasn’t until I started producing longer videos that the business side began to change.

Long-form videos also required a different approach.

Holding someone’s attention for 15 or 20 minutes is very different from getting someone to stop scrolling for a few seconds.

The more viewers watched to the end and engaged with my long-form content, the more YouTube recommended my videos.

That recommendation cycle created a ripple effect across the channel.

As the long-form videos gained traction, they also helped drive attention to my Shorts, which hadn’t been getting much traction from YouTube on their own.

One thing I’ve realized is that access to AI tools isn’t what separates successful creators from everyone else.

Anyone can sign up for the software and generate images or videos.

The difficult part is figuring out what people actually want to watch and then building a process that consistently delivers it.

Advertisement