Blog

Still, Ryan: 1,500 prompts for one AI animation scene

Kakao released Still, Ryan, an animated short made with ChatGPT Images: sketches, about 1,500 prompts for one scene, hand retouching. The method.

Frank Houbre

Frank HoubreFounder of Imaginode

3 min read

What Kakao released

Kakao, the Korean messaging giant, put "Still, Ryan" (그래도, 라이언) online on November 11, 2026: an animated short of about five minutes starring Ryan, the lion of its Kakao Friends, made with OpenAI.

According to Kakao's press release, the film adapts a webtoon of the same name and marks the character's tenth anniversary. It is directed by **Han Ji-won**, director of the Netflix animated film *Lost in Starlight*, with **Kim Young-jun** as AI supervisor. The images were produced with **ChatGPT Images**.

The film came out on Ryan's Instagram, the Kakao Friends YouTube channel and OpenAI's Instagram and YouTube accounts. Before that, Kakao screened it on November 8 in a Seoul cinema, with a talk with the director and an exhibition on how it was made.

Kakao's press release, with the film's poster.Source: kakaocorp.com

The method: sketch, tries, retouch

The interesting part is not the film, it is how it was made. Nobody typed a sentence and got a scene.

According to Kakao, the team started from **sketches of the scenes** and the prompts they wanted, generated with ChatGPT Images, then the director and the supervisor **retouched the results by hand**. For a single scene, about **1,500 prompts** were used to pin down the staging and the look.

The AI supervisor says it plainly, according to Digital Today: even when a generated image looked good, if it did not carry the intended emotion, the team did it again. And it was the creators' judgment and touch, when picking and retouching, that set the final quality.

For you, 1,500 prompts is not a target. It is an honest order of magnitude: a scene that holds on screen takes many tries, and sorting matters as much as generating.

Digital Today sums up the release in English.Source: digitaltoday.co.kr

Apply the method to your projects

You don't need a studio to reuse the logic of Still, Ryan. Three habits are enough.

**Start from a sketch, even a rough one.** It sets the framing and where the characters stand before the first generation. The prompt then describes the light, the texture, the emotion. Our guide to writing a prompt that works covers that part.

**Keep every try of a scene in one place**, with its prompt. When you reach version twenty, you should be able to go back to version six without searching. On the Imaginode canvas, every generation stays next to the others, with its prompt.

**Retouch instead of regenerating** when the image is almost right. A broken hand can be fixed, a good composition does not always come back. To go from stills to motion afterwards, our piece on AI image interpolation explains how to link two shots.

The Still, Ryan method in three steps. Diagram by Imaginode.Source: kakaocorp.com

What we don't know

The release describes the method, not the production figures.

Kakao does not say how long the film took, how many people worked on it, or how the stills were animated: the release talks about images generated then retouched, without detailing the motion step. The 1,500 prompts refer to "a scene", not the whole film.

How we analysed this

Written on November 16, 2026 from Kakao's press release of November 10, 2026 (in Korean) and Digital Today's article of November 10, read on November 16. The screenshots come from those two pages, taken the same day.

We had no access to the production files. The figure of 1,500 prompts is Kakao's. Disclosure: Imaginode offers GPT Image 2.6 among its image models, from the same provider as ChatGPT Images.

Want to try this on a real canvas?

Everything described in this article runs on Imaginode, in your browser.

Get started