 |
|
You are not logged in! F.A.Q Log in Register |
 |
|
 |
 Now online (1)big ...and 91 guests Last 5 registered Oplandisks nothingstar N_loop yipe foxtrotromeo Browse members... |
 |
Members 8025Messages 2608923 Today 17 Topics 127234 |
|
|||||||||
| |||||||||
|
|
|
||||||||
|
|
|||||||||
| |||||||||
|
|
1. download every one of these videos 2. spend a month tagging, sorting them all, transcribing the text 3. feed the tagged dataset into AI. separate components do the commercial text, the audio, the video 4. start a 24/7 GPT2 generative pizza commercial channel and cut a sponsorship deal with a pizza company 5. order a pizza |
||||||||
|
|
|||||||||
| |||||||||
|
|
separate components, no, that won't do. you need to have a a second layer, where there are more networks trained on how the sales pitch voiceover interacts with the cornball skits |
||||||||
|
|
|||||||||
| |||||||||
|
|
if you watch pizza commercials for an hour, and you close your eyes; see ghostly images of pepperoni floating in the darkness... hear the vague murmur of a SOMETHING for just SOME NINENTY NINE... the tetris effect, i've heard it called... that's all the GPT-2 stuff can really do. generate tetris blocks like your brain does if you play tetris for five hours and then go to splash some water on your face. in my experience, they're quite nebulous. a piece sort of looks like a particular tetris piece, then i focus on it, and it disappears. obviously, this does not add up to generating video of someone playing tetris... but, if you treat it as a basic modular building block, and carefully structure it... this train of thought is stupidly absurd and i love it. |
||||||||
|
|
|||||||||
| |||||||||
|
|
oh you did a triangle too with that dong or just citing |
||||||||
|
|
|||||||||
|
Messageboard index
|
|||||||||
