Live data from Hacker News

Seedance 1.0

seed.bytedance.com

61–70 of 127 posts

Re: Seedance 1.0

#61

I look forward to a day when capabilities like this are trivial and boring to the average person. When my phone (locally) will be able to generate a fully voice acted 24 episode series anime on a whim for a meme with my group chat. It's astounding what we can do now, but will be completely ignorable before we know it, which is equally wild.

Who is going to spend time watching those episodes if content is so easy to make? Everyone will be busy watching their own generated content.

Re: Seedance 1.0

#62
post #57

Earlier quoted context omitted.

Will it learn that I don’t like ads?

It will figure out how to make ads that you do watch. If a generated video is good enough, people will watch it even though it’s an ad.

I remain skeptical that there is enough overlap between videos I’d want to watch and videos that make me buy stuff. Conceivably there could be ads I’d want to watch even though they don’t work on me. A subscription model seems more likely in my case, though.

Re: Seedance 1.0

#63

Has the realism of AI already caught on to that of animated CGI movies? I assume that an expert in CGI can point out obvious flaws in these outputs. But I wonder if it is possible to fix those details by prompting it to change only specific segments. There is also the question of how much compute/money they are spending per second of output, compared to a high-budget Hollywood CGI.

Given how bad regular (non-animated) CGI often is in Hollywood movies nowadays, I don’t think the bar is that high.

Change management will indeed be interesting.

Re: Seedance 1.0

#64
post #38

I look forward to a day when capabilities like this are trivial and boring to the average person. When my phone (locally) will be able to generate a fully voice acted 24 episode series anime on a whim for a meme with my group chat. It's astounding what we can do now, but will be completely ignorable before we know it, which is equally wild.

More content will be made in a single month than all of human history up to this point. No more Disney-fication, no more Marvel / Star Wars "mass media slop". We'll have media that caters to people's long-tail interests. If you have a passion for Egyptology and Atlantis, you'll be able to watch a steampunk adventure about the Egyptians waging war with the Atlanteans. But perhaps with the serious tone of "The Wire". T…

Yay, I can't wait for entertainment to stop being soulless corporate sequels and reboots and derivatives, and now with an LLM I can now get soulless machine generated sequels and reboots and derivatives!

Re: Seedance 1.0

#65
post #62

Earlier quoted context omitted.

It will figure out how to make ads that you do watch. If a generated video is good enough, people will watch it even though it’s an ad.

I remain skeptical that there is enough overlap between videos I’d want to watch and videos that make me buy stuff. Conceivably there could be ads I’d want to watch even though they don’t work on me. A subscription model seems more likely in my case, though.

The ads will be subliminal so you won't even know that you're watching an ad.

Re: Seedance 1.0

#66

I feel so bad for the next generation who will never have watched man made movies, they will not be able to tell whether something is junk or not because there will be no baseline.

I dont think they care as long as the content is good. Even memes popular at that demographic are AI generated today.

Re: Seedance 1.0

#67
post #13

Only light skinned people on the video examples. Ethnic diversity and accuracy used to be a problem with the models of the past. I wonder how the model would excel at prompts grokking at that.

[flagged]

Re: Seedance 1.0

#68
post #45
post #43

Earlier quoted context omitted.

> No more Disney-fication, no more Marvel / Star Wars "mass media slop". We'll have media that caters to people's long-tail interests. Lol. Where do you think these models will get the source data for these amazing generations? Also, an average person's "long tail" is incredibly boring (I know mine is). You will need someone to create a next Breaking Bad, or a next Discworld, or... Your slop-generating machine will n…

> Where do you think these models will get the source data for these amazing generations? Doesn't matter one bit, just like mp3 and torrents didn't stop digital music. If you care, there are already "ethical models", there are synthetic datasets, and soon rights holders will be licensing training data. But it literally doesn't matter. The entire world will be on board soon. IP holders included. > Your slop-generating…

> Doesn't matter one bit, just like mp3 and torrents didn't stop digital music.

mp3 and torrents are storage and distributions mechanisms, not production ones. This is an inappropriate analogy.

Re: Seedance 1.0

#69
post #49

really coo, but wheres the sound? i'd expect that they'd have built in the sound model since its gonna look like SOTA for video, VEO3 is great for video but the audios what knocks it out of the park

I work on AI solutions for a major video streaming company, and the problem with VEO3 is that it doesn't have any consistency between prompts. E.g. I can not upload a reference image of what a character looks like, and if I say in one video "the old priest bends down" and in the next video "the old priests picks up the coin", the priest will look very different between shots.

Veo3 does support image to video, so what you can do is create an image that is the start of a scene, and then use that to generate the actual scene. Unfortunately, Veo3 is really bad at this. I expect this will improve over time.

Although I'm not super excited about this Seedance model personally, I do really like that it focuses on consistency between shots. I hope this puts pressure on increased performance from Veo3 in that regard.

Re: Seedance 1.0

#70
post #38

I look forward to a day when capabilities like this are trivial and boring to the average person. When my phone (locally) will be able to generate a fully voice acted 24 episode series anime on a whim for a meme with my group chat. It's astounding what we can do now, but will be completely ignorable before we know it, which is equally wild.

More content will be made in a single month than all of human history up to this point. No more Disney-fication, no more Marvel / Star Wars "mass media slop". We'll have media that caters to people's long-tail interests. If you have a passion for Egyptology and Atlantis, you'll be able to watch a steampunk adventure about the Egyptians waging war with the Atlanteans. But perhaps with the serious tone of "The Wire". T…

Maybe we can finally get shows/movies that are faithful to the books instead of whatever slop we normally get (Looking at you WoT). Obviously a book != a script (like script for a show/movie, not a code script) but with LLMs I feel confident it could generate a script that could be reviewed directly or review the final video then go back and tweak the script as needed ("Ahh crap, it misunderstood the 'you' in that sentence and the wrong character just spoke in the generated video, let me fix that and re-gen that scene").
Post reply on HN