AI Video Tools Are Exploding. These Are the Best

00:23:13
https://www.youtube.com/watch?v=qZM9pHKjlBE

概要

TLDRThe video discusses the current state and future of AI video tools, highlighting some of the most advanced and user-friendly platforms available. It starts with Runway Gen 3, praised for its high-quality text-to-video capabilities, ideal for creating dynamic title sequences with detailed movement and transformation effects. Luma Labs' Dream Machine is spotlighted for its excellent image-to-video transformations using key frames, allowing for artistic scene transitions. The speaker shifts focus to LTX Studio, which enables substantial creative control, allowing users to build short films with prompts and extensive customization of scenes and characters. Additionally, platforms like Kaa are mentioned for their abstract, visually unique animation styles, suitable for creative projects rather than realistic depictions. The video also touches on lip-syncing technologies, mentioning tools like Hedra for creating expressive avatars. Open-source projects are acknowledged, such as Comfy UI and AnimateDiff, which offer complex customization but require more technical knowledge. Finally, Cling, another high-quality video platform, is discussed, although access is difficult without a Chinese phone number and the waitlist is long. Throughout, the speaker emphasizes the fun and creative potential of these tools, encouraging viewers to explore new ways to leverage AI in video production. The video concludes with a reminder of the rapid advancements in AI, urging viewers to stay updated with innovative platforms like FuturePedia.

収穫

  • 📽️ Runway Gen 3 excels in creating dynamic text-to-video sequences, especially title screens.
  • 🎨 Luma Labs' Dream Machine shines in image-to-video transformations with key frames.
  • 🎞️ LTX Studio offers extensive control for creating short films quickly.
  • 🌈 Kaa provides unique, abstract morphing animations, ideal for creative uses.
  • 🗣️ Hedra and Live Portrait are useful tools for expressive lip-syncing.
  • 🔓 Open-source tools offer deep customization, requiring technical setup.
  • 🌐 Cling has high-quality outputs but is challenging to access internationally.
  • 🎬 AI tools can enhance creative video production beyond traditional methods.
  • 📈 AI video capabilities continue to improve, providing new creative avenues.
  • 👥 Engaging with platforms like FuturePedia keeps users informed on AI innovations.

タイムライン

  • 00:00:00 - 00:05:00

    The speaker highlights the exciting developments in AI video tools, particularly focusing on Runway gen 3, which is currently the best text-to-video model. They showcase its effectiveness in generating amazing title sequences with examples, and provide tips on effectively using Gen 3 by following prompt structures offered in its guide. The speaker shares a personal experience experimenting with prompts, discussing the cases of both success and challenges, emphasizing that while results can be impressive, they are often cherry-picked.

  • 00:05:00 - 00:10:00

    The speaker compares Runway with Luma Labs’ Dream Machine, noting that while Runway excels in text-to-video, Luma shines in transforming images to videos, especially with keyframes. They demonstrate how starting and ending frames can be used to create transitions in videos, discussing logical transformations and showing examples reminiscent of Pixar movies. The speaker describes the creative potential and limitations, sharing some long-transition examples from online communities and mentioning a similar Chinese tool named Cling, but notes challenges in accessing it without a Chinese phone number.

  • 00:10:00 - 00:15:00

    They introduce LTX Studio, emphasizing its control and speed. It can produce a short film from a script or simple prompt with customizable styles and consistent characters. They demonstrate the process of developing a short film script, selecting styles, and using the platform’s editing features. The speaker discusses the benefits of LTX's style reference feature, presenting how static images can be transformed into videos with customizable motion. They express satisfaction with the platform’s flexibility, ability to export projects to Premiere, and the capability to pitch them, sharing user-generated content as well.

  • 00:15:00 - 00:23:13

    The speaker discusses Korea's platform focused on abstract and creative video transformations rather than realism, which provides features like a creative upscaler. They use visuals like bioluminescent creatures to demonstrate keyframe-based transformations and highlight customization settings. The speech concludes with examples showing the unique styles available, noting its suitability for artistic projects like music videos. Additionally, they present comparisons of upscale results, showcasing Korea's capabilities and distinct style offerings, and conclude by giving a nod to open-source communities and their foundational efforts in AI advancements.

もっと見る

マインドマップ

Mind Map

よくある質問

  • What is the best text-to-video AI model mentioned in the video?

    Runway Gen 3 is highlighted as the best text-to-video model available.

  • Which tool is praised for its image-to-video capabilities?

    Luma Labs' Dream Machine is noted for its exceptional image-to-video capabilities, especially when using key frames.

  • What is an advantage of using LTX Studio for video creation?

    LTX Studio allows for significant control over video creation with fast speeds and the ability to build short films from scratch or using prompts.

  • How does Kaa differ from the other AI tools mentioned?

    Kaa focuses on creating abstract, morphing animations that are more creative and less focused on realism.

  • What are some open-source tools mentioned that offer similar capabilities to commercial platforms?

    ComfyUI and AnimateDiff are noted as open-source tools that provide extensive customization and control.

  • Is the Cling platform accessible to users outside China?

    Access to the Cling platform is challenging without a Chinese phone number, and the waitlist is very long.

  • Can AI video tools replicate non-human characters effectively?

    Some tools struggle with non-human characters, particularly those with features far removed from human characteristics.

  • What is the primary focus of the Hedra lip-syncing tool?

    Hedra provides expressive talking avatars that are free to use with limitations on the daily usage allowance.

  • What is the main strength of Luma Labs' Dream Machine?

    It excels in image-to-video transformations and works particularly well when integrating key frames.

  • What's a common challenge while using AI video tools like Runway?

    AI video tools often require fine-tuning prompts and may produce varied results, needing multiple attempts for optimal output.

ビデオをもっと見る

AIを活用したYouTubeの無料動画要約に即アクセス!
字幕
en
オートスクロール:
  • 00:00:00
    I've been experimenting with AI video
  • 00:00:01
    tools and covering them on this channel
  • 00:00:03
    for a long time and right now is the
  • 00:00:06
    most exciting and fun time in AI video
  • 00:00:09
    yet Runway and Luma Labs have been in
  • 00:00:10
    the spotlight and I'll cover some
  • 00:00:12
    immediate real world use cases for them
  • 00:00:14
    but my personal favorite tool is
  • 00:00:16
    actually not one of those making the
  • 00:00:17
    headlines I'll get to that a little
  • 00:00:18
    later plus some amazing lip syncing
  • 00:00:20
    tools and some open source models
  • 00:00:22
    there's been a lot of advancements I've
  • 00:00:24
    got to start with Runway gen 3 since
  • 00:00:26
    it's been dominating my timeline it is
  • 00:00:28
    the best textto video model available to
  • 00:00:30
    use right now something that's
  • 00:00:31
    particularly useful with Gen 3 is text
  • 00:00:33
    like these title sequences it is amazing
  • 00:00:35
    at this I'll show a few of my favorite
  • 00:00:37
    examples I've
  • 00:00:39
    [Music]
  • 00:00:40
    seen this is such a perfect title
  • 00:00:43
    sequence for a cooking
  • 00:00:45
    channel this one also showcases how good
  • 00:00:48
    it is at fluid simulation the physics in
  • 00:00:50
    a lot of these is really good adding
  • 00:00:52
    some sound design would take these to
  • 00:00:53
    the next level for sure and be
  • 00:00:55
    production ready title sequences so I'll
  • 00:00:57
    try one for future pedia I'm logged in
  • 00:00:59
    already and here's a whole Suite of
  • 00:01:00
    image and video tools that Runway offers
  • 00:01:02
    but for Gen 3 I'll just click right here
  • 00:01:05
    then type in a prompt a title screen
  • 00:01:07
    with Dynamic movement the scene starts
  • 00:01:09
    with intricate neon circuitry patterns
  • 00:01:11
    that light up and move on a dark
  • 00:01:13
    background suddenly the circuits
  • 00:01:14
    converge and form the word future pedia
  • 00:01:16
    with a glowing pulsating effect then I
  • 00:01:18
    can choose between 5 seconds and 10
  • 00:01:20
    seconds I'll lower it to five then click
  • 00:01:22
    generate that's all there is to it right
  • 00:01:24
    now this is how that came
  • 00:01:27
    back and that looks amazing since
  • 00:01:31
    futurepedia is a longer word I have
  • 00:01:33
    gotten some misspellings but it gets it
  • 00:01:34
    right A lot of the time this prompt
  • 00:01:37
    structure has worked really well so I
  • 00:01:38
    used one of the sample prompts in the
  • 00:01:40
    Gen 3 prompting guide and modified it I
  • 00:01:42
    modified it a lot really just kept the
  • 00:01:43
    basic structure so I'd recommend trying
  • 00:01:45
    to use the prompt structure they give
  • 00:01:46
    here when you're starting out camera
  • 00:01:48
    movement establishing scene then
  • 00:01:50
    additional details hopefully that will
  • 00:01:51
    help you cut down on rolls they do also
  • 00:01:53
    have a lot of solid keywords in this
  • 00:01:55
    guide to help give you some inspiration
  • 00:02:00
    another thing it's really good at is
  • 00:02:01
    transforming between scenes this is
  • 00:02:03
    another prompt you can modify for
  • 00:02:05
    yourself that comes back with great
  • 00:02:06
    results pretty
  • 00:02:08
    [Music]
  • 00:02:12
    consistently I will copy that and paste
  • 00:02:15
    it in
  • 00:02:16
    here I'll change this to a wormhole and
  • 00:02:19
    into an alien civilization
  • 00:02:21
    [Music]
  • 00:02:30
    that's not bad with some rerolls and
  • 00:02:32
    modifying The Prompt I'm sure I could
  • 00:02:33
    get a better generation here's a couple
  • 00:02:35
    other examples using this same prompt
  • 00:02:37
    [Music]
  • 00:02:43
    [Music]
  • 00:02:48
    [Music]
  • 00:02:54
    idea now I've also had a lot of Misses
  • 00:02:57
    with runaway 2 that's the biggest caveat
  • 00:02:59
    here the amazing results you see posted
  • 00:03:01
    are definitely cherry-picked that's the
  • 00:03:03
    case with all AI video models including
  • 00:03:05
    Sora but Runway can do some amazing
  • 00:03:07
    things especially if you have a lot of
  • 00:03:08
    credit here's a few more of my favorite
  • 00:03:10
    Runway examples this was really hard to
  • 00:03:12
    narrow down
  • 00:03:13
    [Music]
  • 00:03:33
    stop
  • 00:03:36
    [Music]
  • 00:04:11
    the best image to video right now is
  • 00:04:13
    dream machine from Luma Labs you can
  • 00:04:15
    also do text to video which gets some
  • 00:04:16
    good results but where Luma shines is
  • 00:04:18
    image to video and even better than that
  • 00:04:20
    is with key frames I'll show a couple
  • 00:04:21
    straight image to video examples first
  • 00:04:23
    it's easy to use upload an image then
  • 00:04:26
    add a prompt volcano erupting contained
  • 00:04:29
    within a drinking glass surrounded by
  • 00:04:31
    peaceful
  • 00:04:33
    Tranquility this was the original image
  • 00:04:35
    then here's the result and that looks
  • 00:04:38
    perfect I'll show a few more straight
  • 00:04:40
    image to video examples then we'll move
  • 00:04:41
    on to key
  • 00:04:53
    frames these are all within one or two
  • 00:04:55
    tries it does weird things on occasion
  • 00:04:57
    like this astronaut grows in extra
  • 00:05:00
    finger or here's a weirder
  • 00:05:05
    one or how about a much weirder
  • 00:05:10
    one most of the time they do look good
  • 00:05:12
    without having to do too many rolls the
  • 00:05:15
    next step is adding an ending frame and
  • 00:05:16
    you can start doing some really cool
  • 00:05:17
    stuff you upload a starting frame then
  • 00:05:20
    also upload an ending frame now add a
  • 00:05:23
    prompt for it to use to create the whole
  • 00:05:24
    clip in between those frames this works
  • 00:05:26
    most consistently between things that
  • 00:05:28
    are logical to transition between if
  • 00:05:30
    they're not then try to be extra
  • 00:05:31
    descriptive of what you want to happen
  • 00:05:33
    in between it works great for
  • 00:05:35
    transformation shots that are in the
  • 00:05:36
    same or a similar
  • 00:05:39
    scene but you can also do more out there
  • 00:05:45
    [Music]
  • 00:05:55
    transitions this one looks like it could
  • 00:05:57
    legitimately be a Pixar movie
  • 00:06:01
    you can extend that out by just taking
  • 00:06:03
    your ending frame and using that as the
  • 00:06:05
    starting frame for a new generation then
  • 00:06:07
    adding a new ending frame you can do
  • 00:06:09
    really long sequences this way there's
  • 00:06:10
    been these long mashups of memes that
  • 00:06:12
    people have been posting which are fine
  • 00:06:15
    I am definitely more interested in the
  • 00:06:16
    creative stuff this with this girl and
  • 00:06:18
    the Wolf is one of the coolest I've seen
  • 00:06:20
    using that
  • 00:07:04
    it's amazing to keep that all consistent
  • 00:07:05
    and there's a ton of possibilities for
  • 00:07:07
    this you can do this for free but the
  • 00:07:09
    generations take a long time like
  • 00:07:11
    multiple hours when I tried before
  • 00:07:12
    signing up paying will get you to the
  • 00:07:14
    front of the line and a lot more
  • 00:07:16
    Generations there's one more out of
  • 00:07:17
    China that's similar in quality to these
  • 00:07:19
    two called cling however there is a huge
  • 00:07:21
    weight list for it and the process to
  • 00:07:23
    sign up is difficult without a Chinese
  • 00:07:25
    phone number so I'm going to cover that
  • 00:07:26
    one at the end moving on to LTX Studio
  • 00:07:29
    they they have the most control out of
  • 00:07:31
    any of these platforms and the fastest
  • 00:07:33
    speed they can build out an entire short
  • 00:07:35
    film in a few minutes how it works is
  • 00:07:37
    you can start from scratch input an
  • 00:07:39
    entire script or start from a simple
  • 00:07:40
    prompt I'll go with the prompt option
  • 00:07:42
    for this example I will paste this in
  • 00:07:45
    it's about a futuristic city controlled
  • 00:07:46
    by AI entities and a hacker that can
  • 00:07:49
    communicate with them he joins a
  • 00:07:50
    resistance they battle all that good
  • 00:07:52
    stuff hit next and it starts
  • 00:07:55
    working it will show the basic story and
  • 00:07:58
    the cast then you can select a style so
  • 00:08:00
    this is Cinematic but let's say I want
  • 00:08:02
    an anime I can direct it more with words
  • 00:08:05
    as well I'll add
  • 00:08:07
    cyberpunk there's all sorts of other
  • 00:08:10
    styles here to
  • 00:08:15
    try I like comic book it will use these
  • 00:08:18
    as consistent characters throughout the
  • 00:08:20
    story which is amazing because that's
  • 00:08:22
    usually a struggle and I can change
  • 00:08:24
    anything about them that I want I think
  • 00:08:25
    this will be better without a last name
  • 00:08:27
    I could change the essence appearance
  • 00:08:29
    the clothes I can test out different
  • 00:08:31
    voices if I want as an organizer I start
  • 00:08:34
    from where the world is as it is not as
  • 00:08:37
    I would like it to be you can even face
  • 00:08:39
    swap these with a face you upload I want
  • 00:08:42
    to change the title to how about Neon
  • 00:08:44
    Flux see how many people know that
  • 00:08:45
    reference to the anime not the movie
  • 00:08:47
    they made from it this looks good here
  • 00:08:49
    so I'll click Start and just from that
  • 00:08:51
    it builds out an entire short film with
  • 00:08:53
    fully editable
  • 00:08:55
    scenes this whole thing was generated in
  • 00:08:58
    less time than it took either of the
  • 00:08:59
    others to generate a single clip and I
  • 00:09:01
    want to show the style reference feature
  • 00:09:02
    again now that this is open because it's
  • 00:09:04
    even cooler to do it in here I'll try
  • 00:09:06
    uploading my own style reference so this
  • 00:09:07
    is a mid-journey image when I upload
  • 00:09:10
    that style every shot across every scene
  • 00:09:12
    regenerates in the new style that's
  • 00:09:14
    super helpful and doing it here will
  • 00:09:16
    give you a much better idea of how it
  • 00:09:17
    will look and that looks awesome I'm
  • 00:09:19
    actually going to run with that style
  • 00:09:21
    now these are just static images
  • 00:09:22
    currently I could generate them all into
  • 00:09:24
    videos at once or open each one to
  • 00:09:26
    customize it that gives you the most
  • 00:09:27
    flexibility so I'll do that I can
  • 00:09:29
    regenerate the image or upload my own
  • 00:09:31
    even generate or remove objects from the
  • 00:09:33
    image I'll just get rid of this little
  • 00:09:35
    blob
  • 00:09:36
    [Music]
  • 00:09:38
    here I can choose from four types of
  • 00:09:40
    motion and adjust how much of it most of
  • 00:09:42
    the options are pretty straightforward
  • 00:09:43
    but this one is really amazing when you
  • 00:09:45
    switch to custom then orbit then you can
  • 00:09:47
    have total control over the start and
  • 00:09:49
    end of the shot and it will generate the
  • 00:09:51
    motion in
  • 00:09:52
    [Music]
  • 00:09:56
    between you can also adjust how long the
  • 00:09:59
    clip is I want this to be like a movie
  • 00:10:00
    trailer so 3 seconds will be
  • 00:10:05
    plenty you can also add music sound
  • 00:10:08
    effects it generates a voice over all of
  • 00:10:10
    that can be customized as well if it
  • 00:10:11
    hasn't hit home yet everything can be
  • 00:10:13
    customized I'll actually change the
  • 00:10:14
    voiceover a little because this ended up
  • 00:10:16
    telling the entire story I didn't
  • 00:10:18
    mention in the prompt that it was for a
  • 00:10:19
    trailer I don't want to spoil the ending
  • 00:10:21
    the track it used is awesome so I'll
  • 00:10:22
    keep it that's a lot of the features
  • 00:10:24
    probably enough for this showcase so I'm
  • 00:10:25
    going to go through and generate each
  • 00:10:27
    scene how I want it then come back all
  • 00:10:28
    right finished all the scenes it's
  • 00:10:30
    really nice that you can export all the
  • 00:10:31
    files so I can bring it into premiere
  • 00:10:33
    for more fine grained editing the other
  • 00:10:34
    really cool option is to export it as a
  • 00:10:36
    pitch deck I'll do that as well first
  • 00:10:39
    here's the pitch deck it has the story
  • 00:10:41
    all of the characters and their
  • 00:10:44
    backgrounds the mood board and color
  • 00:10:47
    palette then the full story board people
  • 00:10:49
    have already made some amazing things
  • 00:10:51
    this way where they storyboarded a short
  • 00:10:52
    film with LTX then actually filmed and
  • 00:10:55
    produced the entire thing they posted
  • 00:10:57
    this one called winged on X where this
  • 00:10:59
    turns into this like human head with
  • 00:11:01
    bird wings it's pretty good you should
  • 00:11:03
    go watch
  • 00:11:04
    it then with this premere option it
  • 00:11:07
    gives me the project file that I can
  • 00:11:08
    edit further I already opened that and
  • 00:11:10
    synced things up to the music a little
  • 00:11:11
    more and made just a couple other simple
  • 00:11:13
    edits here is the final result in the
  • 00:11:16
    neon lit City I discovered a secret
  • 00:11:18
    power the ability to communicate with AI
  • 00:11:23
    Spirits joining the resistance I became
  • 00:11:26
    a spirit Guardian fighting alongside
  • 00:11:28
    comrades unique Tech
  • 00:11:32
    Powers we Face a rogue AI threatening
  • 00:11:35
    our City's balance blending ancient
  • 00:11:37
    wisdom with cuttingedge technology as
  • 00:11:39
    the battle intensifies the fate of our
  • 00:11:41
    world hangs in the
  • 00:11:45
    balance I've been having a ton of fun
  • 00:11:47
    with LTX studio and side note I did
  • 00:11:49
    actually meet some people from the team
  • 00:11:51
    at the Imagine AI conference a while
  • 00:11:52
    back at one point they were just giving
  • 00:11:53
    a little demo at their booth and like
  • 00:11:55
    everyone crowded around to watch they
  • 00:11:57
    really stole the show make sure to sign
  • 00:12:00
    up for the weit list they're letting
  • 00:12:01
    lots of people in the link is in the
  • 00:12:02
    description and thank you to LTX for
  • 00:12:04
    sponsoring this video the platform I've
  • 00:12:06
    been having the most fun with out of all
  • 00:12:08
    of these is Korea and with this one you
  • 00:12:09
    can do quite a bit for free it's a lot
  • 00:12:11
    different than the other ones we've
  • 00:12:12
    covered it's more for abstract stuff not
  • 00:12:14
    so much for realism so it focuses more
  • 00:12:17
    on these like trippy morphing type
  • 00:12:18
    animations which I like a lot personally
  • 00:12:21
    I like to use AI to create things that
  • 00:12:23
    would be impossible to create any other
  • 00:12:25
    way to open up entirely new creative
  • 00:12:27
    Avenues rather than replace EX in ones
  • 00:12:29
    so I have a ton of fun in Kaa I'll go up
  • 00:12:31
    to generate then video they also have a
  • 00:12:34
    creative upscaler that's really useful
  • 00:12:36
    I'll show that one next these are the
  • 00:12:37
    three images I want to use they're these
  • 00:12:39
    translucent kind of bioluminescent
  • 00:12:41
    flower jellyfish and dragon I think they
  • 00:12:44
    look cool together so I'll click add key
  • 00:12:46
    frame select the flower then I'll add
  • 00:12:49
    another key frame for the
  • 00:12:51
    jellyfish and another for the dragon you
  • 00:12:55
    can make these longer or shorter I'll
  • 00:12:57
    lengthen it a little bit
  • 00:12:59
    then you can add a text prompt I just
  • 00:13:01
    want these to morph into each other so I
  • 00:13:03
    don't need any longer prompt I'll just
  • 00:13:05
    say what they are then drag the length
  • 00:13:07
    of the prompt and that's where the
  • 00:13:08
    transitions will start at least that's
  • 00:13:10
    where they're supposed to it's not
  • 00:13:11
    perfect there's a couple other settings
  • 00:13:13
    for the aspect ratio motion intensity I
  • 00:13:16
    like it around 60 usually then you can
  • 00:13:18
    switch the looping on or off it defaults
  • 00:13:20
    to on where the end will morph back into
  • 00:13:22
    the first frames so it can just
  • 00:13:23
    endlessly Loop then you have four styles
  • 00:13:25
    to choose from I'll start with film then
  • 00:13:27
    click generate video and I'll generate
  • 00:13:29
    one at each of those for comparison in a
  • 00:13:30
    second this will do a first pass at
  • 00:13:32
    generating then give a preview at a low
  • 00:13:34
    quality that'll give you a good idea
  • 00:13:36
    what it will end up looking like if it's
  • 00:13:38
    not going in the direction you'd like
  • 00:13:39
    you can go to where it says enhancing
  • 00:13:41
    and select click to cancel that will
  • 00:13:43
    save you some generation
  • 00:13:44
    time now I actually think it's moving
  • 00:13:46
    away from the jellyfish too fast so I
  • 00:13:48
    will cancel I'll add another key frame
  • 00:13:50
    and select the jellyfish again so it
  • 00:13:52
    hangs out on that for longer then we'll
  • 00:13:54
    generate again that cancel option is
  • 00:13:56
    especially helpful if you're on the free
  • 00:13:57
    plan since your Generations are more
  • 00:13:59
    limited I am on the basic plan which
  • 00:14:01
    gets plenty of time for me in about a
  • 00:14:02
    minute it's all done and here's what I
  • 00:14:05
    [Music]
  • 00:14:08
    got I think that looks really cool now
  • 00:14:11
    I'll generate with render animation and
  • 00:14:13
    experimental so you can see the
  • 00:14:14
    difference
  • 00:14:18
    [Music]
  • 00:14:33
    I think these are amazing really fun and
  • 00:14:35
    unique here's another quick example I
  • 00:14:37
    did with these
  • 00:14:42
    [Music]
  • 00:14:47
    images I am working on some longer stuff
  • 00:14:50
    with tracks from suo but they're not
  • 00:14:51
    finished yet so here's one more weird
  • 00:14:54
    one I did
  • 00:15:02
    I love this style I know it's not for
  • 00:15:03
    everyone but it's really great for
  • 00:15:04
    something like a music video or an
  • 00:15:06
    opening sequence but these are a couple
  • 00:15:08
    of my favorite examples I found on x
  • 00:15:14
    [Music]
  • 00:15:26
    [Music]
  • 00:15:29
    oh
  • 00:15:34
    [Music]
  • 00:15:41
    [Music]
  • 00:16:00
    now I also want to demo their video
  • 00:16:01
    upscaler this is not like a traditional
  • 00:16:03
    upscaler where it resembles the original
  • 00:16:05
    video as much as possible just at a
  • 00:16:07
    higher resolution it does a creative
  • 00:16:09
    upscale similar to how magnific works
  • 00:16:11
    but for video it stays really close but
  • 00:16:13
    kind of reimagines everything with AI so
  • 00:16:16
    I'll use a clip from LTX video I made
  • 00:16:18
    this one has a face that was just
  • 00:16:19
    completely warped we'll see if it can
  • 00:16:21
    fix
  • 00:16:22
    that I don't need it upscaled that much
  • 00:16:25
    I'll leave that at 1.5 frame rate 30 FPS
  • 00:16:28
    write out a prompt of what it thinks is
  • 00:16:30
    in the video that looks good so I'll
  • 00:16:31
    leave the strength and resemblance at
  • 00:16:33
    the defaults first and start with
  • 00:16:35
    cinematic turn Loop off and enhance this
  • 00:16:39
    took around 30 seconds and here's the
  • 00:16:41
    result you can see that there's a face
  • 00:16:42
    on here now that's pretty amazing it was
  • 00:16:45
    able to fix that I ran one at each of
  • 00:16:47
    the different presets so you can see the
  • 00:16:48
    difference cinematic and render are
  • 00:16:50
    fairly similar then animated makes it
  • 00:16:52
    more of an anime style I'll show that
  • 00:16:54
    with a few clips from that LTX video
  • 00:17:03
    [Music]
  • 00:17:05
    I hope to see them expand that into more
  • 00:17:07
    Styles because it is very useful overall
  • 00:17:09
    I think Kaa is a ton of fun and there's
  • 00:17:11
    some additional options to play around
  • 00:17:13
    with in here as well lip syncing has
  • 00:17:15
    made some big improvements recently
  • 00:17:16
    there's tons of impressive demos that
  • 00:17:18
    have come out that we don't have access
  • 00:17:19
    to use but there's two platforms I want
  • 00:17:21
    to show that we do have access to one
  • 00:17:23
    that's completely free and one that's
  • 00:17:24
    free for five uses per day it's hedra
  • 00:17:27
    and live portrait hedra has some of the
  • 00:17:29
    most expressive talking avatars I've
  • 00:17:31
    seen it's pretty easy to use so either
  • 00:17:32
    generate the audio or upload some of
  • 00:17:34
    your own I'll use the classic Fight Club
  • 00:17:36
    line now I'll upload a character you can
  • 00:17:38
    also generate one here if you need now I
  • 00:17:40
    just generate the video it works pretty
  • 00:17:42
    quickly and here's what I
  • 00:17:45
    got the first rule of Fight Club is you
  • 00:17:48
    do not talk about Fight
  • 00:17:50
    Club second rule of Fight Club is you do
  • 00:17:54
    not talk about Fight
  • 00:17:57
    Club that looks great super expressive
  • 00:18:00
    let's do another one but with a
  • 00:18:01
    non-human character this time I'll use
  • 00:18:03
    the line from
  • 00:18:08
    Taken what I do have are a very
  • 00:18:12
    particular set of skills skills I have
  • 00:18:14
    acquired over a very long career skills
  • 00:18:17
    that make me a nightmare for people like
  • 00:18:21
    you so that one looks great but it can
  • 00:18:23
    definitely struggle with non-human
  • 00:18:25
    characters and the less human they look
  • 00:18:27
    the harder it is a dog just didn't work
  • 00:18:28
    at all or even some of those gnomes
  • 00:18:30
    especially the ones with a beard it was
  • 00:18:31
    having a hard time with you do not talk
  • 00:18:34
    about Fight Club Runway also has a lips
  • 00:18:37
    syncing feature that they implemented it
  • 00:18:38
    works really well very high quality but
  • 00:18:40
    it's much less expressive the first rule
  • 00:18:42
    of Fight Club is you do not talk about
  • 00:18:45
    Fight
  • 00:18:45
    Club second rule of Fight Club is you do
  • 00:18:49
    not talk about Fight
  • 00:18:52
    Club live anime is also amazing but
  • 00:18:55
    takes a different route you upload a
  • 00:18:56
    reference video and it will map that
  • 00:18:58
    onto the Avatar this allows you to have
  • 00:19:00
    more control over the expressiveness
  • 00:19:01
    it's on hugging face so you can use it
  • 00:19:03
    for free upload a source portrait I'll
  • 00:19:05
    give my face a try then upload a driving
  • 00:19:08
    video it works best with a straight on
  • 00:19:09
    shot like the ones they have in these
  • 00:19:11
    examples I'll just use one of those
  • 00:19:12
    here's a pretty classic one although I
  • 00:19:14
    won't be able to use the song that goes
  • 00:19:15
    along with it on YouTube now click
  • 00:19:18
    animate and in under 30 seconds here's
  • 00:19:21
    the
  • 00:19:24
    result that worked pretty perfectly I'll
  • 00:19:26
    show an example that didn't work this is
  • 00:19:28
    a CLI clip from Breaking Bad it shows
  • 00:19:30
    his face the whole time but he does move
  • 00:19:32
    a bit so with that it's just not able to
  • 00:19:34
    map that onto my face at all so there's
  • 00:19:35
    limitations here but you can still do
  • 00:19:37
    some pretty great stuff with it things
  • 00:19:38
    you would never be able to get through
  • 00:19:39
    other methods currently and really quick
  • 00:19:41
    here I want to shout out the open source
  • 00:19:43
    community and the tools and workflows
  • 00:19:45
    they've been developing and pioneering
  • 00:19:47
    those have really been the foundation of
  • 00:19:48
    some of these other tools that will
  • 00:19:49
    mimic or outright take those workflows
  • 00:19:52
    and Implement them into a paid model the
  • 00:19:53
    things you can get from Korea and some
  • 00:19:55
    of these other video models are possible
  • 00:19:57
    with open source tools most notably
  • 00:19:59
    comfy UI and animate diff that method is
  • 00:20:01
    significantly more complicated although
  • 00:20:03
    it offers more customization and control
  • 00:20:05
    the paid platforms tend to offer a nice
  • 00:20:07
    UI and ease of use so far all of these
  • 00:20:09
    examples are from James G not sure if
  • 00:20:12
    that's how you say his last name he has
  • 00:20:13
    so many amazing Creations on Instagram
  • 00:20:15
    I'm only showing short clips from them
  • 00:20:17
    they're all much longer and have music
  • 00:20:19
    to them as well so I have been following
  • 00:20:20
    his work for a while and it is some of
  • 00:20:22
    my
  • 00:20:24
    favorite this marble statue dance was
  • 00:20:26
    one of his really early ones to go viral
  • 00:20:28
    his Instagram is full of amazing stuff
  • 00:20:30
    but I'll show some other people too this
  • 00:20:32
    dancing spider is
  • 00:20:35
    [Music]
  • 00:20:38
    awesome these have all been from driving
  • 00:20:40
    these animations using real footage but
  • 00:20:42
    that's not the only way to use this at
  • 00:20:44
    all this one by Fab dream going through
  • 00:20:46
    the evolution of money was
  • 00:20:52
    [Music]
  • 00:20:53
    great or these weird frog animations
  • 00:20:56
    from boltron I love these
  • 00:21:00
    [Music]
  • 00:21:03
    to learn more about how to do this I am
  • 00:21:05
    not the person to be teaching it but if
  • 00:21:06
    you want to do a deep dive to install
  • 00:21:08
    and run it locally and learn how to do
  • 00:21:09
    all of that I'll have a couple resources
  • 00:21:11
    in the
  • 00:21:13
    description cling is the platform that's
  • 00:21:15
    as good or better than Runway and dream
  • 00:21:17
    machine with text to video and image to
  • 00:21:19
    video but the weight lless is huge and
  • 00:21:21
    the process is more complicated if you
  • 00:21:22
    don't have a Chinese phone number so I'm
  • 00:21:24
    going to link to a thread by tile AI he
  • 00:21:27
    walks through the process of downloading
  • 00:21:28
    a couple apps and getting a burner
  • 00:21:30
    number to use I was able to sign up for
  • 00:21:31
    the wait list using my own number on
  • 00:21:33
    their mobile app but then the new
  • 00:21:34
    website doesn't have an option for a
  • 00:21:36
    non-chinese number to sign in so it is
  • 00:21:38
    probably best to use his method and fair
  • 00:21:40
    warning I signed up for the wait list
  • 00:21:42
    about 2 weeks ago it said there were
  • 00:21:43
    around 500,000 people on it I still
  • 00:21:46
    don't have access and now the wait list
  • 00:21:48
    says it has around a million people on
  • 00:21:50
    it so it is free to use right now but I
  • 00:21:52
    imagine that will change by the time
  • 00:21:53
    people that are just now signing up for
  • 00:21:55
    the wait list get access so it may not
  • 00:21:56
    be worth the hassle that's why I didn't
  • 00:21:58
    want to spend all time going over that
  • 00:21:59
    here follow that thread if you want to
  • 00:22:00
    try but the quality coming out of cing
  • 00:22:03
    is amazing
  • 00:22:07
    [Music]
  • 00:22:42
    we are finally to a point where you can
  • 00:22:44
    use AI to make video that's usable in
  • 00:22:46
    the real world outside of memes not
  • 00:22:48
    usable for everything of course there's
  • 00:22:49
    still plenty of limitations these tools
  • 00:22:51
    have come a very long way and are a ton
  • 00:22:53
    of fun right now if you want to keep up
  • 00:22:54
    with everything in AI make sure to visit
  • 00:22:56
    futurepedia find the best AI tools stay
  • 00:22:59
    up to date with Innovations learn how to
  • 00:23:01
    use it all with curated tutorials
  • 00:23:03
    there's tons of new features on the site
  • 00:23:05
    so check it out if you haven't for a
  • 00:23:06
    while then subscribe so more videos like
  • 00:23:08
    this show up on your homepage thank you
  • 00:23:10
    so much for watching I'll see you in the
  • 00:23:12
    next one
タグ
  • AI video tools
  • Runway Gen 3
  • Luma Labs Dream Machine
  • LTX Studio
  • Kaa animations
  • lip-syncing
  • open-source models
  • creative control
  • title sequences
  • dynamic animations