Short Film GOOSE for Astana AI Film Festival

The short film GOOSE was created by a team of Russian authors specifically for the Astana AI Film Festival. The project is dedicated to the theme of a future worth living in and demonstrates a unique approach to video generation using neural networks.

Project authors: Daniil Dranenko, Elizaveta Kishkimbaeva, Tatyana Aryzhakova, Svetlana Litvinovskaya, Anastasia Fedosova, and Natalya Marshalhkina. The idea belongs to Svetlana Litvinovskaya; the screenplay was written jointly by Tatyana Aryzhakova and Svetlana Litvinovskaya.

The film's concept is built on the psychological aspect of the future. The authors rejected typical cyberpunk images of space and flying cars, focusing instead on human connections. The future is viewed as a moment here and now, rather than as a distant prospect.

The plot revolves around a person standing on the verge of jumping from a bridge. He is saved by an eccentric stranger who drags him into an urban chase. Amidst the chaos, the hero finds what he has been living for. The Goose acts as a full-fledged participant in the events.

Production took about two months. The main difficulty was harmonizing the time of day in scenes generated at different times. A particularly difficult long dialogue scene in an alleyway posed challenges due to Seedance artifacts and color grading issues.

The team faced a number of technical and creative challenges:

• Character emotions required careful tuning to avoid the artificiality characteristic of neural network videos.
• The film is dubbed in French, although none of the authors speak it. Each phrase was checked for semantic accuracy.
• The Goose, as the main character, had to actively participate in every scene, requiring separate consideration for each frame.
• Work was primarily done in Seedance 2.0, which objectively lags behind version 2.5. Some scenes were refined after the release of the new model.

A suite of tools was used to create the film. Video generation was performed in Kling 3.0 and Seedance 2.0/2.5. Locations were created in Soul Cinema, characters in Recraft. Nano Banana pro was used for corrections, while editing and color grading were done in After Effects.

The GOOSE project demonstrates the possibility of creating deep narratives using modern AI tools. The team managed to overcome the technical limitations of older model versions and create a visually cohesive story.