me that was one example I'll show more throughout this video including a full dialogue scene with lip syncing progress in AI video has been accelerating realistic shots consistent characters complex movement and genuine emotion are becoming possible the hard part is putting it all together combining all the tools and techniques so I'll cover in-depth every aspect of the [Music] process you can make some great stuff with straight text to but to maintain consistency across shots and have more control it's much easier to start from images I'll cover a few tips on generating the best images first
then we'll get to all the fun video generating side flux and mid Journey are the leading options for realistic images I'll use both here this is the formula I use for most of my prompts you could call it the 4 s's there's a lot of ways to prompt that will get you the same results but having a system is nice and one quick way to make your shots more cinematic is just adding the word cinematic or cinematic 35mm I do this right at the beginning in the scene part of the prompt which is just a
generic overview of the scene using archetypes and keywords then I'll add more details about the characters then the setting then the style the style section is another spot to make things more cinematic I'll usually find a film stock to use I'll start off making some shots in a Coliseum with a gladiator so I just looked up what film stock they used in the movie Gladiator then I'll add that to the end of each prompt and get pretty consistent results another option if you're using mid journey is using a style reference to add one you just
drag it up to the prompt bar then switch this to the paperclip icon this will use the Aesthetics like the lighting colors and overall Vibe of your image and it will apply that to your generation this could be from a shot you generated or any image you can find a still from a movie you like and use that as the reference you can use this to get more out there and Artistic Styles too it's amazing for that but I'm focusing on cinematic Styles here an impactful way to control your scenes is by using shot types
which I'll usually put right at the beginning of the prompt closeup to focus on a subject's face or small details often used for emotional moments reactions or dramatic emphasis medium shot shows a subject from the waist up to balance detail with context used for dialogue scenes and general character interactions establishing shot is a wide shot of the location to set the scene provides context and orientation for the audience typically used at the beginning of scenes or to transition to new locations low angle shot where the camera looks up at the subject to make the subject
appear powerful dominant or threatening used to emphasize a character strength or importance high angle shot looks down on the subject making them appear small vulnerable or insignificant or sometimes just to give an overview of the scene taking that further is an aerial shot taken from high above like from a drone or helicopter provides Grand scale and overview used for establishing locations or showing large scale events over the shoulder shot is where the camera is positioned behind one character looking at another character or an object creates a sense of connection between them commonly used in dialogue
scenes to show reactions and POV shows the scene from a character's perspective which increases the audience immersion and identification with the character used to reveal what the character sees or to create suspense Dutch angle is when the camera is tilted to one side which creates unease disorientation or tension it's used a lot in horror scenes sometimes to give the feeling something is not quite right other times to add energy and movement use those shot types in your prompts to help control the compositions the feelings you want to invoke and how you want to guide your
narrative and we'll combine all these with camera movement in the video section another really important factor for making short films or even just a trailer or anything with multiple shots is character consistency the way to do this in mid journey is really easy you drag your character up to the top then select the little person icon to use it as a character reference there's also a character weight parameter you can put at the end of your prompt -- CW space then a number between Z and 100 100 tries to match the Face clothing and accessories
zero only matches the face then the varying levels in between those I want to show this one as an example because it demonstrates something mid Journey struggles with small details like the cut on his face to fix that I can open up the editor then in paint to get rid of the ones in the wrong place and add one in the right place I'll submit this to remove the first one now I'll open up one of those go back to the editor now I want to switch this to a single cut like the one on
the original character you can do a lot of other stuff in this editor like zoom out and pan then edit remove or add objects anything that is blank mid Journey will regenerate when you submit this and you can change the prompt to add the new elements you want included HubSpot has put together a complete guide to YouTube for business and it's totally free it breaks down exactly how to achieve specific business goals through YouTube whether you're looking to build brand awareness generate leads or reconnect with existing customers it covers in-depth YouTube content strategies everything from
what gear you actually need to specific content formats that work best for different business goals they show you how to optimize everything from keywords to thumbnails they even include content ideas to get you started plus there's a whole chapter dedicated to everything you need to know about the algorithm there's seven pages just on that so it's a very detailed guide packed with actionable tips grab your free copy of creating a YouTube channel for your business from the link in the description and thank you to HubSpot for this guide and for sponsoring this video character reference
in mid Journey only works well with mid Journey J generated images if you try to use like an image of yourself or other non- mid Journey images it's way worse but you can do that in flux and with flux there's two main options first the easy way that only needs one image both of these are on replicate by the way also both methods do cost money you put a card on file and it will charge based on usage the first method is just under 4 cents per image on replicate search for flux PID open that
up and it should look like this from there you upload your character so here's an image of me now I'll write a prompt just something simple like wizard casting a spell I want the aspect ratio to be 169 the rest you can start at the defaults they're pretty good but you can adjust as needed except all switch to a PNG for the output then it takes around 20 seconds and here's the result that's really good especially with only one image reference the second option involves more steps but you can train the model on more images
so the results are more consistent on replicate right now it's the main model at the top of the explore page or search for this flux Dev Laura model so first write a name for where the model will be stored and I'm going to keep mine as private next find 10 or more photos I have almost no normal photos of me every photo of me in the past four years also has my son in it except for my thumbnails so I'll use those you just compress your images into a zip file and drag it on to
upload then create a trigger word you'll use in your prompts to refer to this character I'll use my name I'll type a photo of Kevin into the auto captions this isn't important I'll increase the rank to 32 which helps to train on complex features the rest is good now create training and that should finish in about 20 minutes that took 22 minutes now I can click run trained model you can also access it anytime from your dashboard type in a quick prompt Kevin is a wizard casting a spell Kevin was the trigger word I'll switch
the aspect ratio to 169 then scroll past all of these I'll just switch it to PNG again I don't like webp then run and in 12 seconds I have this in incredible image of me as a wizard except I don't know that I'd call that a wizard hat but the face is basically perfect you're not always going to have 10 images of a character you want to create to train this on but if you do this is the best option right now and I'm not sure the exact dollar amount and it will vary but mine
was only like a couple dollars to do this whole process right onto the fun part there's a lot of really great image to video tools now I'll show a mix of Runway cling and Mini Max for this they each have different strengths the biggest strength for Runway is speed it is so much faster than all the others it's not even close like it'll take 30 seconds to generate a video on turbo where it could take 5 to 10 minutes with minia Max or even longer with cling for most of these Gladiator shots I was also
getting the best results from Runway and since I needed to generate a lot of shots it was an easy Choice here that's not the case later where we need to generate more emotions here's an example with a manticor Runway kept it fairly coherent when it was opening its mouth the others had a lot of weird inconsistencies and artifacts Minx did a a lot of weird stuff but one example where this was flipped was trying to get the gladiator's fist clenching Runway and cling both really struggled with this one for some reason I got like 10
bad results in a row from Runway but solid results from Minimax and sometimes the best result will still have some morphing or weirdness to it I'll show how to fix that using the next tool one of the most important factors for making your shots more cinematic is camera movement for my prompts I start with the camera movement and shot type then describe what I want to happen in the scene since it's image to video I haven't found it very useful to describe any aspects of scene or colors or anything like that the things we needed
to do to generate the images I only write what parts of the scene need to move or change here's a really good list of some cinematic shot types to use static shot where the camera remains in one fixed position it's a really versatile shot that can be used to create stability or Draw focus without the distraction of camera movement it can also build tension depending on how it's used tilt is where the camera rotates vertically on a fixed axis to reveal tall objects like in a hero shot which creates a sense of awe or feeling
of insignificance tilting down can increase tension or reveal information pan is where the camera rotates horizontally on a fixed axis to reveal new information new characters or follow action it can create a sense of space or connect different elements in a scene a slow pan can build anticipation a fast pan can create urgency handheld is when there's no stabilization on the camera which can represent a character's Viewpoint give a sense of realism or urgency it's great for documentaries bound footage or to create tension similar but a little different is a POV shot where the camera
acts as the eyes of a character you may even be able to see the character's hands or something they're holding these can be very immersive tracking is when the camera moves alongside a moving subject usually maintaining a consistent framing of the subject while the background changes can have a really Dynamic feel it's often used in action sequences or to follow a character's motion dolly in or Dolly out when the camera steadily moves in or out on a subject in to increase intimacy or tension out to reveal context or create emotional distance Dolly Zoom is the
vertigo effect it's a achieved by using Dolly movement combined with zoom in the opposite direction so the subject Remains the Same but the perspective on the environment changes the only generator I found to get these at all is Runway and they're not super consistent but they do work sometimes aerial drone shot to get sweeping Dynamic shots from high angles used for establishing shots revealing Landscapes or following actions from above can create a sense of scale or Freedom another thing you'll want to add into your characters is emotion here's a great demo covering the range of
emotions this is from Kai Turner and I'll generate some of my own in a second he had chat GPT describe the facial movements as if describing how an animator might create that expression that was a really good idea and he did leave the prompt assistant on as well I'll use a few of those keywords in my descriptions video tools have gotten much better at this it's fairly straightforward on how to get this just add the emotions descriptively into your prompt but it can get overlooked now I've actually found Runway to struggle with this minia Max
and cling are both pretty good and generally if you want more complex movements Mini Max is probably the best but that can lead to some morphing as well and I'll get back to the Gladiator shots later with upscaling and sound design but for emotions I have some shots of two people sitting around a table and I used a few different shot types for a short conversation I found Runway to be pretty bad at this Mini Max was really good but I actually used cling for most of these because they have a lip syncing feature built
in it's great that we can generate these videos with emotions but you know their lips are just moving randomly so I'm going to add the lip syncing but first we need the dialogue to use I'll show a cool way to control the emotions and timing in your voices that I feel like gets overlooked this is in 11 Labs which is the best Texas speech generator out there it's been around for a long time there's tons of options for different voices you should be able to find one for whatever character you have then you just type
your dialogue select the voice and run it but if you need someone yelling or crying or angry it's hard to get that with text to speech using the current tools to get that I'll use speech to speech which is under voice changer if you record the dialogue yourself it will transfer the new voice onto your words but retain the same emotion timing and inflection you used so here's my voice it had no mercy no control it was pure destruction now here's that with the new voice it had no mercy no control it was pure destruction
so it mapped it on really well here's how it would sound if I just did straight text to speech it had no mercy no control it was pure destruction now for mapping the speech onto your characters cling and Runway both have lipsyncing features built in I'll show an example from both in Runway click lip sync then you can type your dialogue in here and use one of their voices cing has that option as well but since I have my own audio I'll upload that and generate one part of this school and Runway is this is
also a standalone feature and you can upload any video to lip sync now cing doesn't offer it as a standalone feature it only works with videos you generated in cing but for that it's the same process click lip sync it will identify the face then upload your audio and generate it had no mercy no control it was pure destruction this can work well a lot of the time but you don't have complete control over what happens another option to have more control is live portrait so I'll show an example I did with a viking running
into battle that I generated the lip syncing didn't work well in cling so I used a live portrait and it was amazing I upload the video here then come over to upload a driving video which I filmed myself singing the words to the song then I run that and here's the [Music] result with this you can not only add the lip move ments but also facial expressions that's really useful it is completely free and open source if you have a computer you're able to download it onto which is what I did well also link to
where you can use a web based version that's not free though but for the scene I used a combination of those lip syncing tools then I added some sound effects and music and a few more scenes what was it Mary what did you see it was this blue creature with crazed bulging eyes I'm so sorry how many did it get all all of them it had no mercy no control it was pure destruction that's impossible there's not one cookie left my god let's move on to upscaling there is traditional upscaling to add resolution remove noise
and enhance details by far the best tool for this is topaz it is also the most expensive but it's unparalleled in quality the option that's free is cap cut they have a free video upscaler under magic tools it's really easy to use it's not on the level of topaz but it will add resolution and also some sharpening and den noising it's a free and easy way to enhance the quality a [Music] bit but what I want to focus on here is creative upscaling which we can do in Korea this is under enhance you upload the
video and you can change the sliders depending on what you need there's presets for style in this case I'll be sticking to cinematic then I'll submit and when it comes back it's done a lot to not only make it sharper and higher resolution but it's fixed to the morphing in the face and the artifacts that were in some of the other movements this can be extremely helpful it is not perfect and with faces while it can fix them if you're using a consistent character even with the relevance set to high it may still alter them
too much but I do use this a lot and it has a pretty generous free plan to test it out or just upscale something here and there and of course there's paid options for more usage good sound design changes everything the sound effects and music it makes it more impactful and conveys the emotion I'll show the process for the Gladiator Clips you can use a stock website for this I personally use story blocks there's lots of options including free ones like pixa Bay those are great but you can also generate sound effects and music so
I want to show that really quick I'll start with sound effects I'm using 11 Labs again let's do footsteps on dirt those could work how about a manticor roar those sound good especially if you layer a few of them together generally it's going to be easier to use a stock website that's what I did for most of these but it's pretty cool to be able to generate them and now with music same thing you could use stock music or generate some andso does a really good job with this type of music so I'll add a
good description of the type of music I want Ando generates ridiculously fast here's the first option I'll stop it there it keeps going for a while but that sounds great I'll generate a few more options I want to test out one with strings I also want to test out like a minimal piano track I'll pick the one that works the best to put this all together I use Premiere so I cut the clips down I layered in just a bunch of different sound effects then added some background music over the top and here it is
[Music] if you want to dive deep into creating images I have a full mid-journey masterclass video here that's a good one to watch next and make sure to check out futurepedia to find the best AI tool for any use case save favorites to your profile browse a curated list of AI tutorials and subscribe to the newsletter for AI news and tutorials delivered straight to your inbox thank you so much for watching I'll see you in the next one
