I was in charge of the direction for Hoshimachi Suisei's "Shoot for the Moon."

Directed and handled the compositing for the lyric video of BUMP OF CHICKEN’s “I,” the ending theme for the TV anime My Hero Academia. The video was crafted so that the various “Quirks” featured in the series would color the screen in sync with the lyrics. Packed with direction inspired by the story, it was created to deliver a level of visual impact that goes beyond the usual scope of a lyric video.

Inspired by ramko’s song, I wanted to express in this video the feeling that “if you stay true to yourself and remain as you are, you can suddenly become aware of emotions that are easy to overlook in everyday life,” and that was the concept behind this production. In the MV, Kuromi-chan’s inner conflict is portrayed with a look that drifts between dream and reality. I hope viewers can enjoy the feeling of peeking a little into Kuromi-chan’s heart through the visuals.

We produced the music video for Midnight Grand Orchestra’s “Soliloquy.”

The designs of the main duo, the girl and the radio-shaped robot, were created by Yoshimi Mibiyoshi, and the modeling of both was done by Sonnya Kamisama. I was in charge of the design, modeling, and animation of the other characters and the city in the background during the CG section of the final chorus from 2:43 onward.

***

Terminal Slam Music Video is a Diminished Reality / Mixed Reality music video in which all advertisements in the city are erased or replaced with Squarepusher-related ads. Daito Manabe created this MV, imagining that in the near future, as shown in this music video, AR/MR glasses will make it possible to freely rewrite cityscapes. Using machine learning technology, we performed: - Object Detection (identifying the positions of objects in an image) - Semantic Segmentation (labeling each pixel in an image and categorizing image regions according to their meaning) - Image Inpaint (image restoration) Most advertising billboards in the city change periodically, so with current technology it was difficult for machines to recognize them accurately; therefore, the billboard position data was created manually. In addition, mask information was generated from the analysis data and positional data, and post-processing was performed, such as applying glitch effects only to areas where people or advertisements exist, and applying optical camouflage effects. Furthermore, because the process of extracting advertisement regions could not be accomplished using existing machine learning technologies alone, manual operations were also carried out for elements that could not be realized with the technology available at present. We anticipate that in the near future, all of this will be done automatically.