How We Created Our AI Holiday Card

Jan 29, 2025

A case study on how we created our AI Holiday Card.

Background

At the end of each year, the creative minds at Slow Clap come together to brainstorm a concept for our Holiday Card. It’s our thing—wrapping up the year by showcasing our humor, talent, and passion for video. This last year was no exception! We love brainstorming ideas and trying different styles of videos (👀 check out our LEGO stop motion from 2023).

Approach

We know Artificial Intelligence (AI) is a hot topic of discussion and is increasingly being used in video. So during our initial brainstorm we explored the idea of using AI tools like ChatGPT, Sora, Runway, Midjourney, etc. to help us with the holiday card creation. We were tight on time so we wanted to do a simple image or video to post on our social media. But what’s the fun of just posting an AI image? We wanted to get the team involved and why not showcase their acting capabilities too. Kiante, Dan, and Darissa became the star of this year’s holiday card video along with guest star – ChatGPT.

We also wanted to bring back some characters from previous holiday card videos (everyone remembers Diva Dan from 2022 right?) and so it seemed logical to develop the storyline for the 2024 holiday card around that… boss taking some leisurely time out in the middle of the day, drops a big task on the frazzled production assistant, you get the drill.

We scripted out a short narrative about Kiante, our production assistant, who is tasked to create the holiday card with less than 24 hours notice and turnaround. Dan, the boss, is off to his tennis match and hands over the holiday card for Kiante to lead. Darissa, the co-worker, sees Kiante struggling after a whole night’s work on MacPaint (our play on MS paint) and suggests ChatGPT as a solution to his creative problem.

At the end, we reveal the beautiful masterpiece that AI has created for our holiday card.

Execution

After we locked our script, we went straight to production with a 1-day shoot at our Slow Clap office. We got to shine the spotlight on Kiante, the main character of this video. While Kiante is usually our production assistant, this was his chance to flex his acting skills and charismatic personality on screen. Kiante makes YouTube videos in his spare time and has done some acting in other short films outside of Slow Clap. It was great to see another side of him!

“It was cool to not have to worry too much about the gear and cameras; I just got to focus on being the character. Katy was great at directing me on set too.” - Kiante Marron, Production Assistant

[Left] Dan operates a camera, while Katy directs Kiante on set. [Right] Dan captures close ups of Kiante’s desk.

Our small crew of four pulled double duty, with three members stepping in as actors! Staying resourceful, we used a locked-off camera setup to film the scenes featuring Dan, Darissa, and Kiante, moving the camera setup around as Katy directed. This efficient approach allowed us to shoot everything using just one camera setup.

[Top Left] Locked off camera set up. [Top Right] Dan acting for his scene. The end results of the shots of Dan [Bottom Left] and Darissa [Bottom Right].

“The Holiday Card shoot is always a great opportunity to wear many hats! Darissa, Dan and I wrote the script, and I helped direct and also shoot some of Kiante’s b-roll shots— as we’re a small team, producing internally, we have to pick up as many roles/tasks as we can. Just another day at Slow Clap HQ!” - Katy Bailes, Producer

One of the biggest challenges was creating a seamless transition from day to night to show the passage of time, followed by capturing an authentic nighttime scene in which Kiante was working late into the wee hours…

In the end, the initial day-to-night transition was straightforward, as it naturally occurred with the timing of the shoot (shout out to things taking longer than planned). However, the nighttime setting didn’t appear dark enough to really convey the sense that Kiante was working into the night.

[Left] 600d light shining on a tree right outside of the window. [Right] Camera and light set up for the night scene.

To achieve this, we needed to enhance the atmosphere. We used a 600d light—a powerful light source—positioned just inside the office window and directed it toward a tree outside. This setup mimicked the glow of a streetlight, casting a subtle ambiance across the set. The result perfectly captured the essence of a late-night work session, which was exactly what we needed!

Actual night scene shot after color correction.

To effectively light the night scene with Kiante, we used a few additional techniques. For example, we placed MC lights on Kiante’s monitor to illuminate him while he was seated at the desk. These lights were hidden from the viewer’s perspective and worked perfectly, especially since we had limited space to light Kiante from the front. After we nailed the lighting, Kiante had fun portraying “hysterical late night Kiante”.

Kiante getting in the groove of his creativity.

Post Production

In post-production, the team drew inspiration from earlier Slow Clap holiday card videos to maintain continuity in tone and storytelling. Our editor leaned into playful, plucky string arrangements, which complemented the video’s mischievous and lighthearted tone from the acting.

The video also incorporated graphic animations of Google Mail and ChatGPT interfaces, which were designed in Figma and polished in Adobe After Effects. Nothing unusual here, but if you take a closer look at Kiante’s desktop and inbox, you’ll get to see a taste of his character.

The email scene when Kiante first finds out he is leading the Holiday Card project.

AI Challenges

Utilizing AI was central to this project and that came with many challenges. While our team members are no stranger to playing around with it, we have never used it to generate actual images of specific people before (we’re used to capturing those ourselves 😉). We knew it would take some time to experiment… but man, we didn’t think it would take THAT long…

We started off using ChatGPT to generate illustrations of our team in holiday settings and gave prompts that included baking cookies and ugly sweaters. ChatGPT quickly produced some entertaining and creative illustrations based off of our reference team image and text prompts. However, ChatGPT couldn’t re-create our team members the way we were hoping.

ChatGPT generated images of our “Slow Clap team”.

Clearly, ChatGPT struggled with a lot of things here. First of all, who are all these people?! We only gave reference to seven team members and kept reiterating that to ChatGPT, but these images still had more people than we asked for. It also struggled with finer details like accurate facial features, gender, race, fingers, and correct text.

Every time we tried to edit the prompt, it would produce a completely new image – some parts for better and also other parts for worse. Even using the premium version of ChatGPT, we weren’t confident that it would get us an image that would resemble our team members after hundreds of iterations. This defeated the purpose of our personalized holiday card.

Recognizing these limitations on ChatGPT, we decided to explore another AI image-generation tool called Midjourney. We found that Midjourney allowed us to be more precise in our prompt edits and we utilized the spot editor that lets us keep the initial image as opposed to fully producing a new image. The ability to refine the image outputs allowed for us to have greater control to edit the details when only one or two things needed to be tweaked.

The process of using Midjourney editor to make the illustration full body and with different colored shorts.

Another issue that we couldn’t escape was the difficulty in generating all seven team members in a single image. We realized a group image was too complex for AI to put together, so we decided to isolate each person and tweak the prompts according to each individual. Eventually we got to a point where there was some resemblance to what we actually look like! It’s not perfect, but we’ll take it over the other results we were getting.

“Despite the challenges, we loved delving into the artistic power of AI and are excited to see how it continues to evolve and play a role in our creative processes.” - Darissa Hicks, Production Coordinator

AI Animations

This project marked the team’s first time using Runway, an AI tool that transforms still images into animated videos. To create the animations, we took individual character designs from Midjourney to group together in Adobe Photoshop and then we fed the image into Runway. The results were far from perfect. We had some hilarious outputs with Runway. See for yourself!

A wonky Runway output of the Slow Clap team running on the beach (not used in our final video).

“When I was reviewing the AI outputs from Runway, I almost had a spit-take from laughing so hard. It’s horribly funny! AI still has some more work to do here.” - Beijo Lee Wang, Associate Producer

At first glance, it’s amazing to see AI animate an image until you take a closer look at the details. You’ll see distorted limbs, changing facial features, and odd body movements, etc. We knew we couldn’t count on AI to make us look like Baywatch (it was only a dream). So we embraced the fact that there would be odd things in the final animated clip. We did more iterations with Runway and even changed the engines it ran on for different results.

“AI tools like this might not replace professional post-production anytime soon, but they’re great for adding something fun and quick to a project.” - Jake Richard, Post-Production Manager

AI tools are definitely still a work in progress. It might have been quicker for us to go out and film a scene on the beach with our team than to have AI create it for us. We didn’t have the time to keep tweaking the prompts and hoping AI would give us exactly what we wanted. So we just had fun with the AI experimentation and finally came across a decent enough clip to use for our final video. It almost felt like an easter egg hunt to see what AI did wrong in just a 10 second animated clip. For us, it worked in a humorous way for our narrative. Notice anything off about the final animated clip?

A decent output of the Slow Clap team running on the beach (used in our final video).

Results

The final video had us rolling on the floor laughing. Who knew AI could make our team look so good 😉. If you haven’t already watched our full video, take a look at it below.

All jokes aside, this project was refreshing to all of us here at Slow Clap. Our typical corporate videos or documentary-style work is the heart of what we do here, but these short narrative projects remind us that we are filmmakers at the end of the day (and apparently now actors too).

“It was a great opportunity to flex our creative muscles while still applying the same techniques we use in other projects.”- Alexander Flores, Editor

We published our holiday card video on our LinkedIn page and received attention from ChatGPT and OpenAI on our post! We couldn’t give all the credit to just ChatGPT; the real hero was Midjourney and Runway for the final images and animated video clip.

We hope these AI tools will eventually be good enough to use in our video production workflow, but for now we are still putting in the hours to make it worthwhile. Try it out for yourself and share with us your results. If you need help navigating AI tools and using it for your videos, feel free to reach out to us at Slow Clap. Although we can’t guarantee you will have as much fun with AI as we did.

See Past Holiday videos:

2020: Zoom Holiday Photo

2021: Let's Build a Snowman

2022: A Furry Holiday Tale

2023: Lights, LEGO, and Action!