5 ms·

I'm really sad about how little creative control these tools have. They seem great for creating slop, but pretty useless for creating content that someone would

I'm really sad about how little creative control these tools have. They seem great for creating slop, but pretty useless for creating content that someone would love. I love the potential but text isn't really a great medium for describing artistic vision. All these demos are focused on how easy it makes everything. Easy is great, but if everyone is able to make instant cute cat videos or whatever it just devalues it. I want to see turning a photo into a rigged 3d model, letting the artist animate and then generate the video. This technology could be used to increase creative expression, but instead it's being used to squeeze out creative expression

bonoboTP 1mo ago
Text is not the only input. You can provide 3d block outs with rudimentary animation, annotated images with arrows etc, voice recordings of one person acting out some emotion then mapping that to a different character's voice, other uses of video to video, etc. There could easily be at least some time period of low skilled ugly people acting in approximate but shitty ways in cheap sets just to give an input reference to a model and then describing the differences in text, yielding gorgeous people speaking with prestigious accents doing stuff in fancy locations in the output.
api 1mo ago
That'll change. They'll get as many knobs as e.g. photoshop for image editing. You'll also be able, if not already, to use your own voice to communicate the emotion and tone you want but then have it re-render it in the voice of your choice. The "slop" phase is early AI, like chunky ugly very early 3D graphics where you can see the triangles. I'm already seeing images generated by first-generation diffusion models like Stable Diffusion 1.5 (which is small enough to run on a phone) used ironically as meme generators due to the now-retro silliness of what they generate.

bonoboTP 1mo ago

Text is not the only input. You can provide 3d block outs with rudimentary animation, annotated images with arrows etc, voice recordings of one person acting out some emotion then mapping that to a different character's voice, other uses of video to video, etc. There could easily be at least some time period of low skilled ugly people acting in approximate but shitty ways in cheap sets just to give an input reference to a model and then describing the differences in text, yielding gorgeous people speaking with prestigious accents doing stuff in fancy locations in the output.

api 1mo ago

That'll change. They'll get as many knobs as e.g. photoshop for image editing. You'll also be able, if not already, to use your own voice to communicate the emotion and tone you want but then have it re-render it in the voice of your choice. The "slop" phase is early AI, like chunky ugly very early 3D graphics where you can see the triangles. I'm already seeing images generated by first-generation diffusion models like Stable Diffusion 1.5 (which is small enough to run on a phone) used ironically as meme generators due to the now-retro silliness of what they generate.