Search Results
7/21/2025, 11:57:32 PM
>>716088750
In 10 years maybe and if NVIDIA gives some crumbs for the masses in the form of plentiful VRAM. Translations aside, you already got the Sugoi/MingSHiba dude who just released a free LLM model of his Sugoi translator which translates pretty decently. The biggest issue for MTL is censorship (kinda solved by said LLM running locally on your computer if it's strong enough) and just as important: context. MTL needs hefty VRAM to remember if who a character is and what they're saying. Right now every single MTL still hallucinates and causes characters to switch genders all the time. It also can't remember what happened so eventually it'll get events and details wrong too. But if you're using a text hook and your MTL of choice, it's enough to finish and enjoy VNs I guess.
In 10 years maybe and if NVIDIA gives some crumbs for the masses in the form of plentiful VRAM. Translations aside, you already got the Sugoi/MingSHiba dude who just released a free LLM model of his Sugoi translator which translates pretty decently. The biggest issue for MTL is censorship (kinda solved by said LLM running locally on your computer if it's strong enough) and just as important: context. MTL needs hefty VRAM to remember if who a character is and what they're saying. Right now every single MTL still hallucinates and causes characters to switch genders all the time. It also can't remember what happened so eventually it'll get events and details wrong too. But if you're using a text hook and your MTL of choice, it's enough to finish and enjoy VNs I guess.
7/16/2025, 12:04:51 AM
IGN - my dad works for Nintendo: the article <still unlinked>
Page 1