I experimented with a bunch of different tools in order to make a short film with AI. The results aren't perfect. Great video is still difficult to do with the current generation of AI video tools. Arms and hands still leave a lot to be desired and getting a consistent character is still near impossible but we've come a long way very quickly!

Tools Used:
https://futuretools.link/chatgpt
https://futuretools.link/runway
https://futuretools.link/midjourney
https://futuretools.link/leiapix-com
https://futuretools.link/genmo-ai
https://futuretools.link/kaiber-ai
https://futuretools.link/elevenlabs-io
https://futuretools.link/mubert-com
https://www.blackmagicdesign.com/products/davinciresolve/

🛠️ Explore hundreds of AI Tools: https://futuretools.io/
📰 Weekly Newsletter: https://www.futuretools.io/newsletter
😊 Discord Community: https://futuretools.io/discord
🐤 Follow me on Twitter: https://twitter.com/mreflow
🐺 My personal blog: https://mattwolfe.com/
🌯 Buy me a burrito: https://ko-fi.com/mattwolfe

🍭 My Backgrounds: https://www.futuretools.io/desktop-backgrounds

Music generated by Mubert https://mubert.com/render

#AIVideo #generativeart #aiart

In this video, I want to attempt to create a short film entirely using AI.

I've seen a lot of really cool videos circulating on Twitter and I kind of want to figure out how they made them.

So in this video, we're going to figure it out.

Now, the point of this video is to a show you how you can make a short film using AI tools, but also be to show you how to kind of tie a whole bunch of different tools together because I think that's really where all of these new AI tools get really powerful is when you take this tool and you piece it together with this tool and you sprinkle on a little bit of this tool over here and you combine them all to get something really, really cool.

And that's what I want to demonstrate in this video is just pulling all these pieces together and coming out with something completely unique.

So I'm going to use AI to generate the script.

I'm going to use AI to generate the video itself.

I'm going to use AI to generate the voiceover.

I'm going to use AI to generate the background music.

The entire video from start to finish will be AI.

The only thing that I'm not really going to use AI for is I'm going to pull all the elements together inside of DaVinci Resolve, inside of my video editor.

So the first thing I need is a script to a movie.

Now this 1, I don't want any dialogue in it.

I want it to be a sort of documentary style with a voiceover throughout the whole thing.

And I also need to generate a shot list to go with that voiceover.

So to get the script, the voiceover and the shot list, we're going to use GPT-4.

So here's the initial prompt I'm going to use.

I'd like to make a short film that's between 60 and 90 seconds long.

The concept of the film is that it will be a documentary about an AI robot that saves civilization.

The entire film will have a voiceover narrator that tells the story.

Please generate a script for the narrator to read that shares the story.

So I'm basically telling GPT-4 I don't want any dialogue, just tell a story and it will be in the form of a documentary.

Step 2 after this is that I'm going to have it generate a shot list for me.

Let's go ahead and have it generate our narration script first.

All right, so it generated the entire narration and it even gave us a shot list in the initial prompt.

I didn't even have to ask it for a shot list.

I'm going to have it generate a more detailed shot list now, but it actually already kind of envisioned what the whole film is going to look like right here.

So my next prompt for each of the shots in the above film, please give a much more detailed description of the shot itself.

Include as much detail as possible around the setting and any objects or people in the shot.

So now each of these shots I'm going to get a little more detailed so that I can use them for my prompting later.

Alright, so now we have a much more detailed shot list that I can use for prompting.

I'm going to go ahead and create a new folder on my desktop here, call it AI short film.

That way, any clips that I make, I'm going to save right to that folder there.

So to generate these clips, I'm going to use runways, gen 2, and using gen 2, I'm going to have the AI try to generate each of these shots.

So in Gen 2, in order to generate a shot, we simply type at Gen dash 2, and then we just literally paste our prompt in.

So if I jump back over to our OpenAI here, here's our opening shot, a wide angle view of a dystopian cityscape in ruins with crumbling buildings, shattered glass, and abandoned vehicles strewn across the streets.

Fires are still smoldering in the aftermath of the apocalypse, casting an eerie glow on the thick layers of smoke that hovers above the desolation.

So I'm gonna copy this whole thing and let's see what Gen 2 gives us with that prompt.

Alright so it took about 3 minutes and we have our first generated video.

Now I would say this is a pretty spot-on representation of that prompt here.

Based on our shot list, it needs to be 5 seconds, but that shouldn't be an issue.

When I pull this 1 into DaVinci Resolve to edit it, I can actually slow down the pace slightly and stretch it out to 5 seconds.

So now I'm going to go ahead and do the same thing with our next shot, except this time, our next shot needs to be about 9 seconds long.

So what I'll probably do this time is actually generate 2 or 3 of the exact same shot and butt them up against each other.

A dimly lit underground laboratory filled with computers, wires, and advanced equipment.

A team of scientists and engineers wearing tattered lab coats and protective gear work on an AI robot with a humanoid form.

They are soldering circuits, adjusting mechanical joints, and inputting data into a computer terminal.

Let's just go ahead and start by trying to generate the entire prompt here.

I might need to break this prompt up, but let's see what we get by just plugging in this prompt in its entirety.

First, This 1 only took about 2 minutes this time and let's see what we got.

So we've got a bunch of people working in a laboratory.

I'm not really seeing the humanoid robot.

So maybe I'll add another prompt here and focus in more on the robot with the humanoid form here.

I'm just going to take the second half of this prompt, a team of scientists and engineers wearing tattered lab coats work on an AI robot with a humanoid form, soldering circuits, adjusting mechanical joints.

And let's see if we can get another scene with more of them working on the robot here.

All right, so here's what we got from that prompt.

It doesn't really look like a humanoid robot.

But once again, I'm going to try to steer it a little bit closer.

Luckily, I can use all of these clips strung together to fill out the time that I need for this specific clip.

So this time I'm simplifying the prompt a little bit, a team of scientists and engineers and white lab coats leaning over a table with a life-size humanoid robot on it.

Let's see if I can get closer to what the sort of vision of this shot is.

So I really haven't been able to get Gen 2 to generate the image that I'm looking for of a robot sort of on an operating table.

So I'm going to jump over to Mid Journey and generate a still image of that.

And then we'll kind of do a panning shot in the video of that still image to represent that piece of the script.

We'll grab our same prompt here and let's just plug it into mid journey.

This time, I'm going to go ahead and give it an aspect ratio of 16, 9.

So it fills our screen and let's see what we get from this.

Okay, so I think I finally got what I'm looking for up here in image 1.

We have a bunch of scientists kind of working over a robot here.

So I'm gonna go ahead and upscale that 1.

And now what I'm gonna do is I'm going to follow this exact same process for every single scene from our chat GPT script here.

So far, I've got our opening shot and our second shot.

I'm going to try to generate them with Runway Gen 2 first.

If I can't get Runway Gen 2 to create an image that I really like, I'm going to jump over to mid journey, give it the same prompt and try to get mid journey to generate something that looks like what I want.

So instead of boring you going back and forth and doing every single 1, I'm going to go and generate all of the shots real quick and then we'll jump back and complete the process.

Okay, so I've gone ahead and generated all of the shots in the shot list here.

You can see I've got a folder with a ton of different files in it.

We have images here of different scenes from the script here and then I've got a whole bunch of different videos that I generated for different parts of the video.

Now, some of these images here, because they're static, I want to give them a little bit of an animation.

And there's also 1 shot that I didn't generate that.

I think a different tool is going to generate better though.

I'm going to walk through those right now.

In case you're wondering, Gen 2 is currently in a closed beta.

They are slowly letting more people in and I believe it'll be rolling out publicly pretty soon.

If you don't have access to Gen 2, Check out this video that I made called actual AI text to video is finally here.

That teaches you how to use model scope, which is a free version of what you essentially can do with gen 2.

The only downside is it does add little shutter stock watermarks inside of the video, but you can use that and it does something very similar to what Gen 2 does completely free.

Now 1 of the shots that I wasn't really able to get a good version of using Gen 2 was this 1 here a time lapse of the city slowly coming back to life, showing a succession of events such as buildings being reconstructed, plants growing through the rubble, and people returning to populate the urban landscape.

So, I'm going to jump into Mid Journey and see if I can generate a starting image.

I'm going to use this prompt from our shot list here of a dystopian cityscape and ruins.

I'm going to paste this into Midjourney and use this as a starting shot for my time lapse.

All right, so we have some images here of a destroyed city.

I think this top left 1 looks the best, so I'm going to go ahead and upscale that 1.

And then I'm going to jump into this Genmo tool, which you can find at alpha.genmo.ai.

And I'm going to upload the image we just created and use this as my starting image here.

And we have some settings that we can do.

So it's asking for my edit and what I'm going to do is I'm going to jump back to my prompt here and you can see we've got a time lapse of the city slowly coming back to life.

Restore the city to a clean, vibrant, bustling town with happy people and colorful plant life.

I'll go ahead and leave the exclusions there.

So let's see the time lapse asks for specifically 9 seconds.

So we'll go ahead and leave it at 9 seconds and let's go ahead and make the video and see what we get out of it.

Look at that it actually kind of nailed it on the first try.

You can see it really quickly change from this dystopian city to this colorful beautiful city with happy people walking around.

I'm going to go ahead and download the video here and we've got our time-lapse for that scene.

Inside my folder with all my video assets and image assets, some of these are still shot and I want to make some of these still shots a little more dynamic, a little more animated, make them feel a little more video-like.

And there's a couple of ways I can do that.

I can use a tool like Leia Pix which sort of adds more depth to the image and animates it so that it kind of looks more 3D.

And then I can use a tool like Kyber which adds some interesting effects to it.

So I'm going to go through some of these images and run them through LeiaPix, run them through Kyber, and have a bunch of variations and options that I can then use in my final video.

Let's start with this 1 of the robot looking at the screen.

You can see how it kind of adds that 3D effect to it.

I can change the animation, slow it down a little bit by changing it to 6 seconds here, speeding up by changing it to 1 second.

I think I want somewhere in the middle here.

Let's go to, yeah, let's go ahead and save it as about 4 and a half seconds.

Animation style, You can go horizontal where it sort of goes back and forth a wide circle, a normal circle, a tall circle, a vertical or perspective, which is a newer 1 that I actually haven't seen yet.

Let's go ahead and leave it on perspective.

But if it is on perspective, I think I want to slow it down even more.

We can change the amount of motion to more or less.

Put it right in the middle at regular or put it at more.

We can make the focus point real close, center the focus point, or make the focus point farther away Which just kind of changes the look.

Now I can click share and save it as an mp4 And we'll go ahead and download this with the rest of our video files.

Alright, let's go ahead and upload another image here Let's see how it looks with 2 people in the image.

Let's go ahead and do a sort of horizontal animation style.

I think that looks cool for this image here.

I think I like the focus point as far on this 1.

And I'll go ahead and download this image.

I'm gonna go ahead and repeat this process with all the rest of my images.

So I have a pool of images using Leopix that I could potentially use in my video as well.

And I'm just going to kind of fast forward through that part, but pretty much going to do the same process with each 1 of my images that I have available here.

So I've gone ahead and used Leopix to convert all my images into something with a little bit more depth.

And now I want to use a tool called Kyber and run some of my images through here just to give another kind of cool cinematic, colorful effect to some of my still images.

So let's go ahead and click on create video here.

Let's grab this image of a robot planting some plants here and let's continue to our prompt and we'll type an AI robot in the style of realistic video.

Let's just see what happens when we try that.

Let's go ahead and leave it at 8 seconds.

I can always speed it up or slow it down when I edit it.

Let's just kind of leave everything at its defaults and see where it gets us and let's click generate.

You can see It starts with my initial image and it shows various types of robots planting plants.

And I think this will look really cool during that scene of my video.

So we can go ahead and upscale the video here.

And it says it's going to take about 6 minutes.

So while this is upscaling, let's go ahead and work on our dialogue.

So we already have the entire dialogue generated by chat GPT here.

Now to record the dialogue, I'm going to use 11 labs, 11 labs, you get a certain amount of free credits every month, and I'm just going to use free credits and it should do the trick for a pretty realistic sounding voice.

So I'm going to create a brand new unique voice for this video.

I can do that by coming up to voice lab here and clicking on add voice and then clicking on voice design.

Let's go ahead and make it a male voice, middle-aged, American accent.

I'm gonna go ahead and set this you know somewhere in the middle and let's generate and see what this voice sounds like.

First we thought the PC was a calculator then we found out how to turn numbers into letters and we thought it was a typewriter.

So let's go ahead and see what happens when I bring the accent strength all the way up and see how it changes the voice.

First we thought the PC was a calculator.

Then we found out how to turn numbers into letters and we thought it was a typewriter.

I think that deeper voice is exactly what I'm looking for, for this sort of dystopian video that I'm making here.

So let's go ahead and click use voice and let's just call it deep male voice.

Now that we've got this new voice, let's go ahead and use it.

And I'm going to grab all of my dialogue here.

I'm going to copy and paste this first sentence here.

In the darkest hour of humanity when all hope seemed lost a beacon of light emerged from the ashes.

All right let's mess with the voice settings that was a little too robotic.

Let's bring the stability down quite a bit and let's test this again.

If you add these little ellipses at the end of a text chunk here it will actually add a little bit of a pause.

I think that's kind of 1 of the downsides of 11 labs.

It's a little bit hard to control the pace of the voice, but it sounds really good.

So I'm going to go ahead and add the rest of our narration in here and just have it record all in 1 long chunk.

And then I'll sort of break it up inside of the editing a little bit.

So there is going to be some manual work when I edit it together.

In the darkest hour of humanity, when all hope seemed lost, a beacon of light emerged from the ashes desperate to save civilization, a group of scientists and Engineers worked tirelessly to create the ultimate AI robot with advanced intelligence and an unwavering sense of duty.

This AI robot, known as the Guardian, was designed to restore balance and rebuild our broken world.

The Guardian traversed the wastelands, bringing aid and hope to the survivors who, against all odds, still clung to life.

Together with humanity, the Guardian forged a new path, working hand in hand to rekindle the once extinguished flame of progress.

Through knowledge and unity, the Guardian nurtured a generation destined to rebuild and reshape the future of our species.

Over time, the once devastated world began to heal as both humans and AI worked in harmony to create a new era of prosperity.

All right, so that sounds pretty good to me.

All right, now that I've generated the audio here, If I come up here to history, you can see the 1 that I just created.

So I'm going to go ahead and select this 1, click download selected.

And now I've got the MP3 of the entire audio here.

Now, 1 last step, I want to add some background music to the whole thing.

Moobert is 1 of my favorite sites to generate audio tracks from, because you can actually use it for free as long as you give attribution.

So I'm going to go ahead and log in and create a new audio with it.

We want it to be sort of like a dark, let's set like a serious mood here.

Cause if we remember our film should go to about 1 minute and 20 seconds.

So let's go ahead and just save this as 1 minute and 20 seconds and with a tense serious mood let's generate the track and see what we get out of it.

That's almost a little too poppy for for.

So let's go ahead and download this 1, added it to my downloads folders.

It's still sort of prepping it for download and we're ready.

It asks where you're going to publish it.

So I'm going to go ahead and put my YouTube channel and then I will go ahead and also put my Twitter channel because I'll likely put it on both and we'll go ahead and click continue and we'll agree and download.

Alright so now I've got a folder with all of my files in it.

We've got our voiceover, we've got our background music, and then we've got all of our little clips to go along with it.

Now it's time to just assemble it all together, and to do that I use DaVinci Resolve.

You can use DaVinci Resolve completely for free.

Just go to blackmagicdesign.com slash products slash DaVinci Resolve to find it, and I'm going to create a new project here.

I'm going to pull in all of these assets that I just created here.

I'm going to start by dropping my audio onto the timeline so we can listen back to it.

In the darkest hour of humanity, when all hope seemed lost, A beacon of light emerged from the ashes.

And then we've got our background music so I can drag this behind it here.

I like to set background music at about a negative, let's go negative 20 here.

In the darkest hour of humanity when all hope seemed lost, a beacon of light emerged from the ashes.

Desperate to save civilization, a group of scientists and..."

So now I'm just going to start lining up the perfect clips based on what ChatGPT gave me.

See all Matt Wolfe transcripts on Youtube

Make A Movie with AI: It's Crazy What We Can Do!