AI Images and Video Achieve Photo-Realism
Key Points
- Flux Pro now produces 4K AI‑generated images that are virtually indistinguishable from real photos, raising both creative possibilities and misinformation concerns.
- Luma AI’s Dream Machine delivers short‑form AI video of near‑professional quality with improved character persistence, marking a leap comparable to the current state of large‑language models for short text.
- The speaker draws a parallel: AI excels at short‑form content (images, video, text) but still lags far behind human creators on long‑form narratives like novels or feature‑length movies.
- Slev Agents, founded by former Google and Stripe staff, secured $56 million to build an “operating system” for AI agents, signaling strong investor confidence in foundational agent infrastructure.
Full Transcript
# AI Images and Video Achieve Photo-Realism **Source:** [https://www.youtube.com/watch?v=h7HFAQK-jxk](https://www.youtube.com/watch?v=h7HFAQK-jxk) **Duration:** 00:04:35 ## Summary - Flux Pro now produces 4K AI‑generated images that are virtually indistinguishable from real photos, raising both creative possibilities and misinformation concerns. - Luma AI’s Dream Machine delivers short‑form AI video of near‑professional quality with improved character persistence, marking a leap comparable to the current state of large‑language models for short text. - The speaker draws a parallel: AI excels at short‑form content (images, video, text) but still lags far behind human creators on long‑form narratives like novels or feature‑length movies. - Slev Agents, founded by former Google and Stripe staff, secured $56 million to build an “operating system” for AI agents, signaling strong investor confidence in foundational agent infrastructure. ## Sections - [00:00:00](https://www.youtube.com/watch?v=h7HFAQK-jxk&t=0s) **AI Achieves Near‑Photorealistic Images & Video** - The segment highlights flux Pro’s 4K photorealistic image generation and Luma AI’s Dream Machine delivering virtually indistinguishable short‑form video, while noting both the breakthrough potential and the accompanying misinformation risks. ## Full Transcript
four pieces of AI news today two of them
are actually on the image and video
front which I don't tend to talk about
as much but it's seen huge steps forward
between flux Pro which is the first
thing I want to call out and Luma ai's
dream machine which is around video you
have essentially the jump to
undetectable image quality on video and
on photos for AI driven content so that
means if I tell the AI to make a photo
on flux Pro in
4k it's reasonable that 98 99% of people
are not going to notice the difference
versus an actual photograph taken of a
real scene now if I put flying
hippopotamus is in there people are
probably going to know its AI but if I
make it a intentionally realistic
looking scene no one can
tell and that matters because anytime
you get to a spot where it becomes
undistinguishable versus reality you
become a conduit for misinformation and
so tools like this super useful super
helpful but they're going to have a dark
side as well the other major image one
is the dream machine released by Luma AI
I am just so impressed at the way video
has evolved I remember a year ago when I
was looking at AI video and I was just
kind of rolling my eyes because it
wasn't close but with dream machine
we're now at a point where short form
video is something that is almost
indistinguishable from actual video shot
with professional cameras by
videographers and we're making huge
strides on character persistence so the
idea that a character's face image
likeness is something that could be
persistent from frame to frame and you
have a stable sense of character that
was also something that we didn't get a
year ago and we're now making real
strides on so this is just a periodic
update to basically call out that we are
in a place where AID driven uh video and
image generation
is on par with what you get with the
physical camera in the real world except
in the case of longform video and what's
interesting is this is very similar to
where we're at right now with large
language models and text where short
form content and factual content with an
large language model is on par with what
you get with all but the best writers
but long form narrative like stories or
novels isn't anywhere close and I think
that's super interesting okay okay
moving on to the next piece of news
beyond the image world uh a company
called
slev agents got $56 million in funding
they're founded by ex googlers ex stripe
folks and their goal is to build the
operating system for agents and I say
that like I don't report all of the
different funding rounds but I think
this one's significant because the
leadership involved and because of their
focus on building essentially the
infrastructure layer for agents I think
it's a really smart play and I expect
that they will do well and I think the
size of the raise reflects the
confidence of the investor community
that this is a good spot to park cash in
hopes of return so slev agents I would
expect more from them
soon finally uh the rumor is that
Google's Gemini uh llm is going to offer
a service called code base upload in
December where you going to be able to
upload a thousand files and up to 100
100 uh megabytes I believe of pure
code I don't know we'll have to see kind
of how that goes rumors are rumors but
my sense is this is a step in the
direction of a longer context window and
the idea is that we want more and more
to have the llm operative across the
entire codebase now I know some code
bases get far far far larger than this
and uploading will not be the way to
handle that but I do think getting to a
point where an llm can reliably work
across a thousand different files is a
big step toward the idea of near
infinite context Windows which is
something that execs uh certainly at
Microsoft and some other places have
been hinting at in 2025 so we will see
but it's it's an early sign from Google
that we're starting to see those context
Windows expand all right well that's
your news for today got the images you
got the video try and not do
disinformation guys um and yeah we'll
have to see how how the future unfolds
and what tomorrow holds