Behind the scenes: How I created this AI-generated video

This article describes my creative process for my first image-to-video project—yes, the first, but not the last. Using a combination of ChatGPT, Midjourney, and RunwayML, this project was enjoyable and generated many new ideas for future productions.

I signed up for RunwayML about a year ago to test the image-to-video feature. A colleague and I had worked on creating some animated characters for a branding project. While working, we came up with the idea that bringing our characters to life would be advantageous by creating AI-generated videos of our images. However, the result could have been better, as our characters changed appearance during the short videos.

Time passed, and I almost forgot about Gen Ai image-to-video because I thought it was not quite there yet.

Fast-forward one year, I found some generative AI videos in my LinkedIn feed. This time, they were much better quality, and the characters in the videos maintained the same appearance. So I logged into my account on Runway, and wow, it has improved a lot in the past year! 

This time, I wanted to test whether I could create a short intro to a video, and this is how I did that. 

ChatGPT as my co-director

I wanted to do a short test video, a reflective narrative that could create a certain mood for the video without over-committing to a clear message or storyline. I asked ChatGPT for a narrative:

My prompt: "I want to create a 20-second video with a reflective, philosophical tone. The scene should be set in a bustling city, capturing the contrast between the fast-paced world around and a girl sitting on a bench, lost in thought."

That's where the conversations started, and about 10 minutes later, ChatGPT delivered the scene and script for the narrator:

ChatGPT output: "The film opens with an aerial view of a bustling city at dusk. The sky is streaked with hues of orange and purple as city lights flicker into life. The camera slowly zooms in on a lone woman sitting on a park bench, gazing up at the sky, lost in thought."

Narrator (soft, philosophical voice):

There's a fleeting moment in each day—a quiet pause between what was and what is yet to come—when time seems to hold its breath. And in that space, we wonder: What does it all mean? This brief life we walk is so fragile yet so full of possibilities.

Wow, this can be good, I thought...

Midjourney as my photographer

For the video, I needed the following three images, which I generated in Midjourney:

My prompt: "Ultra-realistic helicopter image above a vibrant city at dawn, 2024, ultra-high-definition."

My prompt: "Hyper-realistic Cinematic mid-shot of a young woman sitting alone on a contemporary park bench, gazing at the sky, surrounded by blurred city lights, 2024, high-resolution, dawn lighting."

My prompt: "Hyper-realistic cinematic mid-shot of a close-up of a young woman's face, sitting on a contemporary park bench, with warm twilight lighting, lost in thought, soft bokeh city lights in the background, 2024, ultra-high-definition, clean, sharp focus on the face."

Runway as my filmmaker

The last step was to add the images in RunwayML and describe the camera movements. I used the image-to-video feature to create the motion and the text-to-audio feature to create the narrator's voice reading the script. 

I also used Runway's video editor, which has an easy-to-use interface but is robust and offers many options. 

Here is the result. 

I think this was an excellent first version. Next, I'll start experimenting with Runway Act-One, which allows you to generate expressive characters.

If you have a creative idea that needs to be brought to life—a short video, a dynamic visual story, or something unique—I’d love to hear about it and see if we can help you. 

Feel free to book a meeting to discuss your next project!