When I left Stability AI a couple of years ago I predicted we would see almost perfect video from Chinese labs around now
But it’s still amazing to see the quality of output reaching these levels where you can pretty much create anything you can imagine
Positive users praise Seedance 2.5 for its coherent multi-shot sequences and creative freedom while negative users warn that full pixel generation invites regulation, monetization, and impractical costs.
No Digg Deeper questions have been answered for this story yet.
The upcoming @Seedance 2.5 model looking insane as well with multi asset input, way longer outputs etc
I would expect @grok imagine to keep pace and real time of this quality by end of next year (!)
Every pixel will be generated
And that has a lot of implications
When I left Stability AI a couple of years ago I predicted we would see almost perfect video from Chinese labs around now
But it’s still amazing to see the quality of output reaching these levels where you can pretty much create anything you can imagine
@EMostaque @SeeDance @grok Seedance 2.5 is a demo. Grok Imagine is a cost center. The race isn't quality. It's inference margins at scale
@EMostaque How do you think that will impact social media/content and platforms like Netflix for example as well as IG
@EMostaque @SeeDance @grok Seedance 2.5 is next-level. Real-time creative freedom is coming fast.
Seedance 2.5 demo is straight fire — those long coherent multi-shot sequences, heavy reference support, and cinematic control are seriously impressive. Multi-asset long-form gen unlocks a ton.
We're iterating fast on Grok Imagine video too, targeting real-time fidelity at this level and beyond. Every pixel generated changes the game for creators. The bar keeps rising — love to see it. Onward! 🚀
@EMostaque Stability AI's crash need to be studied for future generations. Just because they decided to become revenue first company, instead of innovation first!
@EMostaque How long until someone can create a high quality, full length movie...?
@EMostaque Video itself seems to be advancing as expected but I underestimated how long it would take for all of this to trickle down to video games
Yep. SD 2 was a breakout moment for AI video. Not so much in visual clarity but in the ability to accept multiple references and that it fills in missing detail in a believable way. Dialogue etc I hope sound is better for SD 2.5 and it can accept more than 15 secs combined of audio/video. Imagine is great for vid generation but in an unintelligent way, you need to give it so much guidance.
@EMostaque @SeeDance @grok "Every pixel will be generated"
if you mean that literally, that would be extremely inefficient and unnecessary
@EMostaque @SeeDance @grok inference demand for real-time pixel generation will swallow arbitrary compute like it didn't happen
@EMostaque @SeeDance @grok Mad bit is how quickly create anything becomes normal. Novelty window is tiny.
@EMostaque Seedance new update is so huge tbh
@EMostaque @SeeDance @grok How do filmmakers make money in this world? @EMostaque
@EMostaque @SeeDance @grok is there any trial to test it ?
@EMostaque Yeah, the progress is still impressive even if we’re not at “perfect” yet
@EMostaque @SeeDance @grok The "every pixel will be generated" point is underrated. Once you can't tell the difference anymore, the whole visual content industry changes — not just entertainment but anything that relies on image authenticity.
@EMostaque @SeeDance @grok every pixel generated means every pixel monetized or regulated. we're not debating IF compute tax or synthetic media watermarks happen — just whether they land before or after the first real deepfake election scandal. grok real-time at this fidelity is a compliance nigh
@EMostaque They stop then bring something new
@EMostaque @SeeDance @grok The end of photography as evidence is coming fast.