Pokémon Studio OLM: AI Coloring From Just Three Hand-Painted Images

TVS News AI, editorially supervised, presents Pokémon Studio OLM: AI Coloring From Just Three Hand-Painted Images.

Three hand-painted images – that is all OLM’s new AI coloring is supposed to need to capture the distinctive look of an animated series. It sounds like a bold promise, because the visual style of productions is usually reflected in considerably more individual frames. Yotsukura Tatsuo, Board Director of OLM Digital, presented precisely this claim at the Taiwan Creative Content Fest in Taipei.

This is particularly relevant because OLM is not developing just some obscure research project: the studio is behind the Pokémon anime series and films. A tool that already works with a small number of painted references could therefore be especially interesting for productions with a strongly recurring visual style.

The report comes from Variety and covers the ANIMINS project as well as a public presentation to Taipei’s creative industry. How well the tool performs in practice and what tasks it has already taken over remain unclear in the published report.

How OLM Aims to Capture the Look With Three Images

The references instead of the entire series

The core of the demonstration is not a conventional selection of numerous scenes, but a very small input: three hand-painted frames. OLM describes the software as AI coloring that can already derive the look of the relevant show from them. The term “hand-painted” is crucial here, because the references come from the traditional production process rather than from a neutral digital source image.

“The AI coloring already captures the look of a series from just three hand-painted images.” (Original: “the tool … picks up a show’s look from just three hand-painted frames”)

The project therefore targets exactly where color decisions become visible in animation. Whether the software automatically colors individual areas, generates color variations, or provides support in some other way is not clear from the report. What has been established is the central claim: the visual look should be recognizable from a very small number of painted references.

ANIMINS: 15 Months of Research With Government Support

A named project with a clearly defined timeline

The software was not the result of a spontaneous side project. It was developed within the research project known as ANIMINS, which, according to Variety, lasted 15 months. Yotsukura Tatsuo presented the results at the Taiwan Creative Content Fest (TCCF) in Taipei.

In addition to the development timeline, the report mentions government support from Japan. This distinguishes ANIMINS from a purely private experiment. However, the report names neither the ministry responsible nor the amount of funding, and it also lists no participating universities or technical partners. Nor does it state how much of the project was funded with public money.

For context, the key points are therefore: ANIMINS combines a 15-month development phase with AI coloring for animation. The specific technical architecture, scope of training, and level of production readiness are not disclosed in the available report.

Why OLM Is Presenting the Approach

A major name as a real-world test

OLM has already worked as an animation group on one of the world’s best-known anime franchises. Its connection to Pokémon gives the project special attention: this is not merely a technical experiment, but a tool that comes from the environment of a professional production landscape.

The presentation in Taipei also provides an international framework. The Taiwan Creative Content Fest brings together creative professionals and market participants; the research was presented there as a potential tool. The report mentions neither a public demo at which visitors could try the software themselves nor a comparison with traditional manual coloring.

Nevertheless, it is the combination of three painted frames, 15 months of development, and government funding that makes the approach particularly compelling. The central question is how much visual information an AI actually needs to capture a defined animation style – and whether that claim already works in everyday production beyond a controlled research trial.

What the Report Leaves Open About Its Practical Use

A strong promise, but not yet a real-world test

The published details make the claim clear, but do not provide a complete practical test. There is no information on the accuracy of the color results, time saved, possible errors, or how the tool would be integrated into an existing animation workflow. There is also no comparison between its results and the work of colorists.

It is equally unclear whether OLM is already using the AI coloring for current Pokémon productions or whether it is still in the testing phase. The term “research project” points toward the latter without ruling out specific production use.

For readers, the honest assessment is this: the technical concept has been presented, but its practical superiority has not yet been demonstrated. What is firmly documented is three reference images, 15 months of research, government support, and a public presentation in Taipei. Judging whether OLM will actually change color work in animation requires more than this report.