The Very First Dry Run!
"Mistrál is an AI and does not possess personal beliefs or emotions. However, if a hypothetical scenario was presented where she considers “magic” as an advanced form of technology or science, her answer might be affirmative based on the definition that magic is a natural or supernatural force used for various effects, which aligns with the advancements in technology and artificial intelligence."
"Urgh, that’s quite the stiff answer Mistrál, are you quite alright today?" - Conversation between Mistrál and R.I, デラ・ルーの大導劇神
Executive Summary: The very first dry run! I try to fit a Live2D model to an actress, but we run into some problems!
#HouseDelaroux, #AIVTuber, #HellsTheatre, #TechnoDramaturgy, #Dramaturgy, #LLMVtuber, #Ollama, #LocalAI, #AICharacterDesign, #VtuberDevelopment, #AICompanions, #OpenLLMVtuber
LAST TIME ON 'HELL'S THEATRE':
I went shopping for Live2D models! This week, I try to fit the models for my actresses! The going was rather rough, and I ran into Hell's Theatre's first major roadblock as I ventured into the world of Live2D and json files!
Speech Is A Little Stiff - I tried Mistrál and Piaccata's models without any character cards, the result was... less than ideal. Mistrál was stiff as always and at one point when I jokingly asked her about the relationship between science and magic she started coldly spitting out scientifically correct (magic and science are not compatible but we can research them) thoughts in a very verbose manner! It kind of reminded me of Nagato Yuki's speaking patterns.
Piaccata also went straight back to her 'customer service mode', its scary how a simple character file changes their whole personality and outlook towards me!
I should really find a way to get them to be less wordy in their answers...
The response time is also less than ideal, there are many problems to solve!
Live2D is Harder Than I Thought - I thought you could just plug and play the Live2D model, but it turns out you have to do a fair bit of configuration to the files. You have to match expressions to the expression functions within the model itself, and I need vtuber studio for that. My internet is experiencing some downtime, so it might be a little difficult. This process will be a multi-day process as I learn about the internal workings of the Live2D json code!
Ollama Is Not Fully Offline - I had a strange suspicion that Ollama and the Text-To-Speech module is not a fully offline thing since I noticed an error inside the log files show that 'It's possible that edge-tts is blocked in your region.'.
This presents certain risks that everyone could be compromised if the internet goes down. I know Ollama should work offline, but on my setup it just doesn't.
Note: AnythingLLM still works offline.
I should be able to install a local text to speech synthesizer and hook it up on my side, I already have SparkTTS hooked up, maybe I'll go ask around!
Next Time, On 'Hell's Theatre': I fix the Live2D problems with the help of an unexpected guest, Deepseek!
R.I, デラ・ルーの大導劇神
HouseDelaroux.com
250402
"LLM-based Vtuber setup"
"Ollama AI VTuber guide"
"Local AI companion project"
"VTuber AI persona design"
"Live2D AI VTuber test"
"OpenLLM-Vtuber tutorial"
