Bad Apple / Maçã Ruim

• 

  Artwork processed with AI which was part of this production, but had to be left out.

• 

  A lot of editing and overpainting for each scene, after adding details each the altered part sent into AI and integrated.

• 

  Alternative image: The extra mouths beside her head illustrate throat singing. In one alternative version I sing a passage in that manner.

Inspiration

The original cover and graphics of the chiptune song by Masayoshi Minoshima blew my mind when I did see it for the first time quite some years ago. Mr Masayoshi have gratiously made the song public domain which made it possible to make a cover [as so many other have done.]

What it does

While Tudi is the deuteragonist of the story, this video made it possible to show things from her perspective, this while she also reference the original Lotus Land Story in her usual chaotic and unorthodox manner.

How we built it

The music track came first, and in no less than 5 varieties, the one used here is hybrid AI, and listen carefully and it is clearly possible to tell what parts of the track that use AI as the sound quality drops. Clips have been made according to the script for 2 years and was added to a preliminary script film. Now that AI film clips got the quality I decided to move on and make this music video - where images in some cases were extracted from the older clips and re-generated.

Challenges I ran into

The most major one was that the entire save file in NCH got corrupted the very day before the deadline! So the video had to be remade, and now presented with a number of small details that could have been addressed if time had been available. During the previous production one problem to be solved was to get the figures and effects to move in pace with the music, where I only can claim a partial success by adjusting the speed. As this would have affected mouth movements, lip sync had to be done last. As this was done with Kling, which was a mixed blessing as it use a non-standard 30 FPS, causing frame drops. Their word filter prevented me from adding lip sync to a second of the song, even though the lyrics do not contain anything offensive.

Accomplishments that I'm proud of

That I managed to recreate the video even though the save file had been corrupted the very last day before the deadline.

What we learned

The use of frame interpolation make it possible to adjust the speed of actions and events, I hope to refine this technique further.

What's next for Bad Apple / Maçã Ruim

The full series of 9 episodes are in production and will be finished, this video is an auxilliary or bonus part of the storyline. I have tried the text to speech generator of Elevenlabs which is my choice for some of the voices in the story episodes.

Built With

• audacity
• bandlab
• deepdreamgenerator
• hailou
• kling
• nch
• seadream
• vsdc