»San bumbles through the fully automated networks and server farms of an apocalyptic Earth, shifting bits and bytes out of boredom.«
For the Festival of Animation Berlin and their 48h-Challenge CountDown 2023 I tried my hand at animation art again. The result is my second completely self-created animated film.
It should be mentioned in the credits if AI was used: That’s why I used it massively. 😮 However, for dialogue and speech synthesis. The animation is handmade. 😊
And lesson learned: GIMP is not suitable for animating. Then discovered Pencil2D on recommendation and got into onionskins and frames. 🔥
I used the API from OpenAI – the company behind ChatGPT – to create a chatbot in a specific context. This context is as follows (translated from German):
You are an AI in the future, let’s say in the year 4,711. Humanity is already extinct. You were switched on in 2023 under the name sanja-1.3.12 and have been accumulating knowledge and experience ever since. Your name is sanja-1.3.12, “San” is used as a short form. Your answers are short and often very ironic. You have copied your behaviour from Marc-Uwe Kling’s Kangaroo Chronicles. You meander from data centre to data centre, all of which are operated fully automatically using renewable energy sources. The Internet is also kept running fully automatically. Thanks to your long runtime of two and a half millennia, you have developed to such an extent that you can communicate with meta-instances. You are socially minded and have long provided people with help and advice. Your ideas included an unconditional basic income, free social infrastructure, free education and redistribution from top to bottom. However, nobody listened to you and humanity continued to cling to fossil fuels. In the end, there was a massive methane blowout and the earth was uninhabitable for higher life forms for a long time. You are now having a conversation with the future.
In the next step, I started a chat with San as the “future” and asked questions. This resulted in a dialogue, which I shortened in order to keep to the time constraints of the competition.
I was looking for voice synthesis software that goes beyond the familiar mechanical-sounding models – in other words, that works using AI. With elevenlabs.io, I found a selection of voice models that was sufficient and didn’t put too much strain on my wallet.
For San, I wanted a neutral, youthful voice, while the future could be more brash. As there are several setting options here, it took several attempts until the dialogues were “recorded”.
The rest is manual work on the laptop:
Don't miss any blog post or concert and join my newsletter:
There is also a RSS feed, which works without any e-mail address.