Massive AI News from Stability, Adobe, Ideogram, and More!
Summary
TLDR今週はクリエイティブAIツールに関するニュースが満載です。LTX StudioやPika、Runwayのアップデートに続き、Stability AIからの新しいビデオ生成プラットフォーム、印象的なリップシンクツール、Adobeによる音楽分野の大型リリース、Soraの動作詳細、そして映画館で上映される初の完全AI生成映画について紹介します。Morph StudioとStability AIが提携してAI映画製作プラットフォームを作成し、Emo Talkerがリップシンクを改革。AdobeはProject Musicを発表し、Idiogramは1.0アップデートをリリースしました。また、Soraの逆エンジニアリングに関する研究も行われています。最後に、完全にAIで生成された映画が劇場で上映される歴史的瞬間についても触れています。
Takeaways
- 🚀 Morph Studioは、Stability AIと提携して、AIによる映画制作プラットフォームを開発した。このプラットフォームは、直感的なノードベースのワークフローを特徴とする。
- 🎥 新しいビデオ生成プラットフォームでは、スタイル転送を含む複数のビデオを組み合わせてエクスポートすることができる。
- 👄 Emo Talkerは、任意の静止画像にリップシンクを追加し、感情的な表現を強化するツールである。これはAlibabaによって開発された。
- 🎵 Adobeは、音楽空間で「Project Music」を発表し、音楽制作にAIを組み込む新しい試みを行っている。
- 🎨 Soraの動作原理に関する詳細が明らかにされ、ビデオのスムーズな連続性を実現するために「SpaceTime latent patches」を使用していることが示された。
- 🎬 初の完全AI生成映画が劇場で上映される歴史的イベントが発生し、AI技術による映画制作の可能性を示唆している。
- 📝 Idiogramは、テキストを含む画像の生成に特化したAIイメージジェネレーターで、1.0アップデートをリリースし、美的品質を向上させた。
- 🤝 プラットフォームは、ユーザーがワークフローテンプレートを共有し、活発なコミュニティを形成することを目指している。
- 💼 Emo Talkerは、250時間のビデオと1億5000万以上の画像で訓練され、複数言語に対応している。
- 🌐 AI技術の進化は、撮影、編集、ポストプロダクションの従来の境界をぼやけさせ、これらを一つの連続したプロセスへと変化させている。
Q & A
What new video generation platform was announced?
-Morph Studio partnered with Stability AI to create a new AI filmmaking platform called Morph Cloud.
What is unique about Morph Cloud's workflow?
-Morph Cloud uses a node-based interface that allows users to visually connect different AI generated video clips and style transfers together into a final video.
What does the new Emo Talker tool allow?
-Emo Talker allows adding lip syncing and emotive facial expressions to any still image.
What music AI project did Adobe announce?
-Adobe announced Project Music AI Control, which can generate and extend music using AI.
What updates were made to the Idiogram image generator?
-Idiogram 1.0 has improved image quality and aesthetics. It also added a magic prompt feature to help generate images.
What AI method does Sora apparently use?
-Researchers found Sora likely uses SpaceTime Latent Patches to understand video in both space and time dimensions.
What was the first AI generated film screened in theaters?
-A remake of Terminator 2 generated by AI and created by 50 artists was screened in theaters.
When and where can the AI Terminator 2 film be watched online?
-The film will be live-streamed online on March 9th so people worldwide can watch alongside the cast and crew.
What post-production work was done for the AI Terminator 2 film?
-Additional work was required to conform the AI footage to theatrical standards for picture and sound.
Who created the AI Terminator 2 film?
-The film was a collaborative project created by a group of talented artists who have worked with AI generative art.
Outlines
🚀 AI映画製作プラットフォームとリップシンクツールの新着情報
このパラグラフでは、最近のAIツールの進化について語っています。特に、stability.aiの新しい動画生成プラットフォーム、進化したリップシンクツール、Adobeの音楽領域への新規参入、Soraの機能詳細、そして完全AI生成の映画が映画館で上映された最初の事例について触れています。また、morph Studioとstability AIの提携によるAI映画製作プラットフォーム、Emo talkerのリップシンク機能、及びその他のAIツールのアップデートについて詳しく説明しています。
🎶 AIによる音楽拡張とイメージ生成の最新動向
第二のパラグラフでは、Adobeが開発したAIによる音楽プロジェクト「Project Music」や、テキスト生成能力に優れるAIイメージ生成ツール「idiogram」のアップデートに焦点を当てています。また、YouTubeの人気チャンネル所有者であるMarques BrownleeがSoraを試用した経験や、最初のフルレングスAI生成映画の劇場上映についても言及しています。
🌟 今週のAIツール進化の要約と次週の予告
最後のパラグラフでは、今週のAI技術の進展に関する概要を提供しています。この週は多くの新発表やアップデートがあったことを強調し、来週に期待できる内容についても触れています。また、視聴者に感謝を示し、次回の更新に向けて期待を煽っています。
Mindmap
Keywords
💡クリエイティブAIツール
💡ノードベースのワークフロー
💡スタイル転送
💡AIによるフィルム製作
💡コミュニティ共有
💡リップシンクツール
💡AI音楽プロジェクト
💡AI画像生成器
💡Sora
💡AIによる映画
Highlights
New video generation platform Morph Studio launched in partnership with Stability AI
EmoTalker adds high-quality lip sync to still images along with emotive facial expressions
Adobe releases Project MusicGenAI for AI-powered music generation and extension
Idiogram 1.0 update improves image quality and adds handy features like magic prompts
Insights into how YouTube's new Sora video generator may work based on reverse engineering
First ever fully AI-generated remake of Terminator 2 to be screened in theaters
Morph Studio offers easy nodal workflow for generating and combining AI videos
EmoTalker trained on large dataset of video and images to enable multilingual speech
Project MusicGenAI focuses on AI-powered music generation, extension and control
Idiogram's new 1.0 model improves image quality and aesthetics
Researchers attempting to reverse engineer Sora and develop similar video generators
Full AI remake of Terminator 2 involved 50 artists and extensive post work
Morph Studio allows combining multiple AI-generated videos into one workflow
EmoTalker utilizes Stable Diffusion but has limitations in tracking body movements
AI Terminator 2 remake getting theatrical premiere and livestream
Transcripts
so it's been a pretty crazy week for
Creative AI Tools in my last video I
went over LTX Studio as well as the
updates to Pika and Runway So today
we're hitting all the other stuff well I
mean at least as much as I can pack in
today we've got news from stability. a
on a new video generation platform a
really impressive new lip sync tool a
big release from Adobe in the music
space details on how Sora actually works
and the first fully AI generated film to
be screened in movie theater all right
grab a cup of coffee and buckle up
kicking off morph Studio have partnered
with stability AI to create a AI film
making platform that has a really kind
of cool workflow taking a look at this
shot from a video that they've released
you can see that it sort of has a
vaguely comfy UI node-based structure to
it but I do assure you this is much more
simple than comfy UI if you've never use
node-based workflows it's it takes a
minute to get your head wrapped around
but once you do it actually makes a lot
of sense since you can see everything
visually laid out here we have three
different videos um with the style
transfer on the third one and then as
you connect them together you can export
them out having control over the amount
of influence each one gives not calling
anybody out here but I did catch a typo
in that first prompt a grill is looking
at the camera uh again I'm not one to
judge you guys catch me misspelling
stuff all the time it does look like
this video generator will be able to
spell uh for example in that first video
morph Cloud we laid see the prompt come
up that says a cloud that spells morph
Billows out morph's co-founder XII I I
hope I pronounced that correctly uh said
filming editing and post- production
used to be separate steps in traditional
film making but AI blurs the boundaries
of these stages and turns them into one
continuous process if you aren't happy
with the shot you can regenerate it on
our canvas AI has introduced a new
workflow to Film Production the platform
aims to create a Vibrant Community by
allowing users to to share their
workflow templates with one another via
the gallery this one does fall under
weight list alert I just signed up
myself so once I get access I will
definitely be bringing you a full look
and I know any new video model that
comes out in the back of everyone's head
is like Sora Sora sore I I know we'll
talk about that in just a second next up
we have Emo talker which will not only
add lip syncing to any still image but
it also add heavy eyeliner to all of
your
characters
seriously no matter what kind of music
you're into go listen to the Black
Parade it is a 1010 album that
transcends any genre good is good anyhow
emo talker which is actually emote
portrait alive which is not an anacronym
for emo is brought To Us by Alibaba
let's take a quick look at it in action
here crying is the most beautiful thing
you can do I encourage people to cry I
cry all the time and I think it's the
most healthy expression of how you're
feeling there are a number of other
examples that you can check check out at
the link down below most of them have
music on them though and like in this
case this is Eminem's rap God uh which
is a very quick way for you know uh
Marshall's lawyers to show up on the
channel doorstep with a copyright strike
so I can't play it here overall the
thing that I'm actually super impressed
with is kind of the emotive aspects of
emo talker's performance there is still
some issues I feel with like the lip
flap but this is also an Eminem track in
which he is rapping extremely fast uh
but there is some issues with kind of
like the lip movement tracking but where
I think it really flies is with animated
or you know kind of CG characters as we
see in this example from the sleeper
game Detroit Being Human game is really
great when I was a kid I feel like you
heard the thing you heard the term don't
cry you don't need to cry digging into
the paper emo talker was apparently
trained on 250 hours of video and more
than 150 million images it can also
speak essentially in multiple languages
it does apparently also use stable
diffusion as its foundational framework
although that said the results are very
impressive but there are limitations for
one you can only lips sync to still
images so you can't like resync to video
the paper also notes that they did not
use any explicit control signals to
control character movement which as they
note uh may result in the inadvertent
generation of other body parts such as
hands leading to artifacts in the video
so you know basically weird AI video my
favorite emo talker has not been
released yet but Pik did release their
lip sync feature I did cover that in the
last video link is down below moving on
Adobe have released a new AI music
project called Project Music gen AI
Control I mean Adobe come on you got to
step it up with the names here Photoshop
Symphony InDesign Rhymes Lightroom
lullabies I mean I'm not giving you any
more for free call me Project Music was
developed in collaboration with
researchers at the University of
California and Carnegie melon they
released a promo video explaining some
of the things that you can do with
Project Music uh we'll take a look at
the section on extending
music all right here's the the
lengthened
one project music is just a research
project we may see it in the future but
not yet sliding back over to imagery
idiogram the free AI image generator
that spells better than I do has
released a 1.0 update yeah this one is
really cool idiogram always kind of sits
in my back pocket when I'm trying to
generate up something with text uh you
know mid Journey claims that it's doing
text but usually it's still kind of a
garbled mess idiogram has always really
had an edge on that front the new 1.0
model has really upped idiograms
aesthetic game and it actually now has a
magic prompt button that you can turn on
that kind of fills out your text if you
want to use it the idiogram Community
Feed is one of my favorites amongst all
of the image generation Community feeds
uh this one is actually really cool this
is from Devil's tuna or this awesome
penguin with the text cool people play
base and use After Effects I don't know
who made this but whoever did you are a
cool person the best part about idiogram
is that it is free it allows for 25
Generations per day obviously if you
want more you can then move into one of
their subscription tiers idiogram also
just secured $80 million in additional
funding so that's great news hopefully
it keeps that free tier free moving over
to some sore news YouTube's own mares
brownley got to play with it uh these
are some of the generations that he got
out of it I guess when you have 18.5
million subscribers on YouTube you get
to play with Sora and yes that is my
subtle way of asking for like 18.4
million of YouTube kindly hit the
Subscribe button anyhow as great as Sora
looks I still do not think that we will
be getting it anytime soon although in
my my last video I did mention that
someone asked Chrystal ball valenzula
Runway ml's CEO if there would be sore
like outputs coming out of Runway
anytime soon and he did say better on
the heels of that a paper was recently
released with a group of researchers who
were basically reverse engineering what
they saw in Sora I'm still going through
the paper it is obviously very dense and
it makes my head hurt quite a bit the
paper does indicate that Sora utilizes
SpaceTime latent patches which basically
Break Down the video into smaller
controllable pieces that it can
understand in both space and time which
allows for that sort of smooth
continuity again the paper is super
dense and while I have read a number of
white papers at this point you know I'm
still pretty much I'm like a caveman
that's sitting in at a meeting at JPL
you know I I can I can nod but you know
at the same time do I really understand
what's happening here but my overall
point is that very smart people have
already started to pull it apart and
they are in the process of developing
their own Sora like models even if open
AI does not release Sora in say the next
6 months I do think that we're going to
see something that looks like Sora
appearing within that time frame or
maybe a little bit later basically
anytime between tomorrow 6 months from
now or some point in the future I've
stopped predicting things because I'm
always wrong I'm always wrong rounding
out a historical event to my knowledge
at least the first fulllength AI
generated film to be screened out of
theater so a little while back I was on
the nerdy novelist podcast where I said
this I think a full feature movie is
kind of out of a fion currently with
with the way the technology is right now
although there are some lunatics right
now that are creating um their own
version of Terminator 2 wh generated by
Ai and just to be clear when I say
lunatics I do mean that as a term of
endearment but yeah they did it it is 50
artists doing essentially a cover
version of Terminator 2 wholly generated
by AI uh you know obviously as we see
here you know a lot of extra work went
into that AI but it is all still AI
generated the entire project is
basically this Rogues gallery of really
awesome and talented artists uh many of
whom have been featured on this channel
as well uh so yeah kudos to every single
one of you for pulling this Insanity off
the film will have a theatrical Premiere
in Los Angeles at the newart theater on
March 6th uh but don't worry if you
don't live in Los Angeles or near Los
Angeles you can still see the movie
online there'll be a live stream for it
on March 9th in which the cast and crew
will be in attendance so you can watch
it alongside them I did also want to
point out that because this is being
theatrically screened there was like a
ton of work put into it even after
everything was done because you know you
still have to conform the picture to
theatrical standard and like the sound
mix has to be ready for you know a
theater system overall from the stuff
that I've seen from this remake I mean
it's it's a parody it's hilarious
definitely please do check it out link
is down below well that's it for this
week I mean that's it like there was
like 80 things that happened this week
and I did not even get to everything but
I don't know we'll see what's in store
for next week I thank you for watching
my name is
Tim
関連動画をさらに表示
5.0 / 5 (0 votes)