Generating Video with AI (Studio Sessions with Bear)

[0:00]This is a recording of studio sessions with bear, a live humanitarian AI training led by professor Nick Bear Brown recorded on May 17th, 20125. In this session, you'll see the team explore real tools for creating educational videos with AI covering everything from script writing and voiceovers to image generation and editing. It's a behindthescenes look at how we build content together in real time. Okay. So, what we're going to do is just sort of uh go through the process of making a video right now. Is Nina on? Yep, I'm here. Great. And this is being recorded? Uh, yes. Great. It is being recorded. And so, the first thing I'm

[0:44]Going to do um is make a well, not the first thing. What we're going to do is basic purpose of this is just how do we quickly make videos? Uh, I think probably since people from the Conbon OPT team are here, uh, I'm hoping that means you're going to want to make videos. So, the first video you should probably make is what is this project? Because if you want to sell that project, we need to easily explain it to people. So, the way I would go about that, uh, can I, uh, share? There you go. Sorry about that. I would, and we can do this for any

[1:25]Article or anything. This is part of the reason why we put the stuff up here so it's easy to find stuff. But we'll go to projects. We'll go to the OPT conbon. And what I would start doing is I would just take the whole thing. I can take it like this. Uh, I'm going to have to use claude in here because the northeastern claude thing is sometimes it works in one browser but not another. Um, so I'll have a prompt like write a voiceover script. Okay. So, the first thing I'm going to do next, not the first thing, what I'm going to do next is create a little

[2:41]Folder. And I'm going to call this folder voiceovers. Actually, I'm going to call it scripts. Uh, one of those is going to be a readme. I'll change in a second. I'm going to download this little script here. Let me upload those scripts to the uh project here in this scripts folder and this is sort of the process because what will happen I know for at least for me what will happen is I'll forget where this script is uh if it's in a session what session it's in etc etc All right. Um, and I'm just going to clean out this

[3:53]Quickly. By the way, you can take old photos of yourself and redo them in midjourney. now. Um then put this here and go back here. And this is where your expertise is going to come in because this script may be wrong or it may not reflect it may not reflect the way you want to uh to say something about it. Um and so you're going to have to read the script and edit it. Uh it also helps to have it uh in a voiceover format. So I'm going to do that right now. So I'm going to say remove the

[5:08]Annotation and we're going to put it in two tools and we can give you access to Uh we can give you access to both these tools. You you have access to the suno question. Okay. And so uh you want to listen to the script. You can also record it yourself if you wish. You can also record it and have 11 Labs take your recording and switch it to a different voice if you want. Uh, but we are going to now render the the sound. And usually with these kinds of scripts, you're going to want to mix some AI footage, what we'll call B-roll, with actual footage of the tool. But you guys

[6:15]Have access to the tool. So you can just use the tool and do a screen share. Uh but let's first try it with 11 Labs. Um the spacing and the capitalization will be important. So, for example, in a uh in a voiceover, it's likely to say opt, but if it were spelled like this, the voice over would likely say opt and you'll see it. It will listen to it. It should say opt if it's capitalized, but it may say op. If it's uph, but we have to listen to it first. Again, I think we have 300,000 or more credits here. So, we you can use this as

[7:22]Much as you want. And we'll use another tool, which I tend to prefer called Sunno for this as well. So, this will add a pause to it if I add more of a pause to it if I do that. But I think this should be fine. So, we listen to it. Welcome to the OPT portal project, a cutting edge web app. That's too slow. I made this super slow on purpose for another So, the speed is super slow here. Let's change that. Like I think there's a thing to do it to defaults reset values. So this will reset it. I know. Welcome to the So that

[8:12]Was way too slow. But you can adjust those here. application designed to transform how educational institutions and international students navigate the optional practical training application process. Every year, thousands of international students face a complex documentheavy OPT application process. I'm not in love with that. You can try different voices or just use your own voice and have it either just use your own voice or or do it. But you can play with things. This is the basic process of AI to try and play. Try and play. Try and play. Try and play. Hey Nick, can we ask questions as you go along? Absolutely. Yeah. How do how do you pick

[9:01]A voice? Okay, good question. I would have not shown you that. You go to voices. Oh, got it. Okay. Oh, cool. You have my voices, which means you can you can upload your own recording and create a voice from a clone. So, let's try a different voice. Let's try. Good day, cobbers. My voice is a fair dinkhamy voice. Great for a riper yarn about life down under. You'll be a true blue Aussie. And welcome to the OPT portal project. A cutting edge web application. So, the problem with the voice like that is the the voice will probably distract from your message because the voice is so funny. Yeah. Um,

[9:42]But I did actually create a voice. uh it's in my voices which I think works well for a lot of stuff. It's a female voice which tends to be less aggressive. And so this is a voice, it's not a default 11 lens voice. This is a voice I created. Uh and I created it actually for the Catbot project because I I felt the Catbot, you know, just was just better with the female voice. It's just softer and not as aggressive as opposed to the the voice you or my own voice. And the voices they had I didn't really like. So this is that one. I just I just name the

[10:24]Voices. Just an arbitrary name. Welcome to the OPT portal project, a cuttingedge web application designed to transform how educational institutions and international students navigate the optional practical training application process. You can play with it. I actually like that voice quite a bit. It's clear. There's no mumbling of anything, but it's not like really aggressive or you can also use Indian voices or whatever voices. I tend though to use Suno for this which you also have access to. We'll go to create and I'll just put that in there and I'll use the style spoken word. And I also add poetry to that to give it a more lyrical feel. What I find is with

[11:24]The 11 lamb's voices, they're very percussive and they don't really have most people speak with a rhythm. They don't really have the rhythm. They're very dut welcome to the OPT Porto project kind of thing. Some of them aren't so much, but that's what I hear when I listen to those voices. So instead, I and because of that, I tried. Sono isn't made for this. That isn't made for voiceover. But so many of the voices had that almost robotic hello, I am a, you know, kind of feel that I just said, well, why not try and so I did. And so let's try. With suno, you work a little differently. You

[12:13]Generate a bunch of them. typically five to seven, but you have 10,000 plus credits, you know, and we have multiple Sunno counts. So, we we've never run out of credits, even though we've made hundreds of songs. Each song that we've made had maybe 20 versions of the song. And then it reloads every month. So, I'm just going to create. Another advantage to Sumo is it generates a wave file unlike 11 Labs which generates an OP uh MP3 file. MP3 files aren't as high quality. They're also not stereo. They're mono. Let's listen to a couple. Welcome to the opite portal project. A cutting edge web application. Welcome to

[13:02]The OP portal project. A cutting Welcome to the OPT portal project. A cutting edge web app. Welcome to the OPT portal project. a cutting edge web application designed to transform how educational institutions and international students navigate the optional practical training application process. Every year, thousands of international students face a complex doc. So, we're going to remove that sound if you want to remove the sound. So, I'm going to download this once it finishes. It hasn't finished actually rendering it. It allowed me to play it before it finished. And I would typically play with the voice a little bit more. I might add something like Jamaican accent or

[13:45]Something to it to just give it a little bit more of a flavor and then I'll like it. I forgot which one I think was this one. Heavy OPT application process. Universities struggle with manual tracking system. Well, I would play with that voice now. I I'm not in love with that voice. Again, the the process of of AI is to quickly generate stuff and filter quickly generate filter quickly generate filter and refine. But because the purpose of this is just to show you, so we download the wave file and use the wave file, not the MP3 file, high quality. And then we open it in a free

[14:26]Tool, which is also the best tool. That's the best situation when the best tool is also a free tool called Audacity and we uh open it up and we look at it. The N looks a little funny here. It should tail a bit bit better. So I'm going to zoom in and I'm going to listen to that. Listen to that from about here. This is the end of the speech. So that's just music. So the tendency of Sinnos to add a little bit of background music, which is actually kind of nice, but I'm going to zoom back out. So it may it looks like the actual

[15:21]Speaking may have stopped here. And we can see that from sort of the patterns in the wave. But let's check that. Higher education administration. Yeah. Higher education administration. So, what what we're going to do, so this is where it actually ended. And then it added just some music trailer, which is nice to have a couple seconds, but not 30 seconds of of um So, we are going I'm going the wrong way, am I? So, what we're going to do is it ends here. Administration looks to be about here where we want to end it. So, we just go to edit and we go to remove special trim

[16:29]Audio. And while we have everything highlighted, we're going to change the the the amplitude. So we go to effect volume and compression and amplify. Here it says if we wanted to make it maximal, that is to have the full expression vocal range, we would amplify it by 2.54. But we never want to take it to the full range. Because if if you've ever heard a speaker like vibrate, it's because they've gone to or above the full range. So, it says if we go to 2.54, we'll get a full range, but we don't really want a full range. We want it up there, but we don't want it full.

[17:10]So, I'm going to take that to two. And there we go. And now, because we sort of cut this off, we're going to fade it down. So, we go to effect again, and then we go to fading, and we go to fade out. It looks like I made a little mistake. had kept something I shouldn't have. So, another way of getting rid of stuff is not to trim, but just to cut. So, I'm just going to cut that. And I can see that I could see that it went down, down, down, and then there was a spike. And we don't want that spike. So, we'll listen to it

[17:53]Again. And higher education administration I love that fade. I think I'm going to fade it again. Administration. Okay, that's better. Hey, I have a I have a quick question. What Why are there two tracks? Is one It sound looks like both are voice. Yeah, it's called stereo. So if one is for your right. So if you if you So I'm deaf in one ear. Got it. Yeah, that makes sense. It's stereo. Sometimes music sounds weird because in like a stereo album sometimes they'll play things on the left mic but not the right. I won't hear it if they don't also put it on the right. Got it.

[18:47]Um and if you opened up a MP3 file, you would only see one track. Oh, good. Makes sense. It's not stereo, but this is more sophisticated. You could actually like what often happens is in in an actual album sometimes they'll have the sounds different, but that's problematic for people like me who can't hear in one ear because they're there. I you may not ever have noticed it on an album. I do whenever they do that to an album because if I'm listening to it on headphones, then it literally what is supposed to go in the right ear I don't hear. Um, but you don't really have to worry

[19:24]About that. It just it should be here. You see there just it's the same duplicate. Uh, but it's this is stereo sound. It's also an indication that you have a a better sound. It's not a mono sound. So then we save save it again. Got another question. Yeah. Um, where's the music track? Is it part is it part of the I'll show you that in a second. So, what we're going to do is this is a combination of all the tracks. So, I'm going to save this to I'm just going to do it to the scripts folder. That's where we put this stuff. And so, this this sound is a

[20:03]Combination of all the sounds. So, there are some free tools. The one that I typically use is called fader. And we create uh something called a stem. Um, so what we would do is we upload a song and I'll I'll sign in. So you can get just a free Google account and they give you quite a few fades on the free one. Oh, I guess I didn't have a I didn't have one in Google. I had I guess one on a different one, but I just signed up for Fader right there. Uh, what we're doing is we're going to create a stem. We're going to upload the saw which is

[20:48]Downloads the version the the save version that we trim is in scripts and what it's going to do it's going to do it's going to remove all the instruments and the vocals and put those into different tracks and you don't have to do this this would only be if you only want to separate the voice. So we're going to download the stems. We're going to download them all and we're going to download them in wave format. So, we download the stems and then am I going to move I'm going to move that file to our scripts folder. What is it called? Fader. Oh, am I in

[21:55]Downloads? Thought I downloaded it. Oh, here we go. and I'm going to move that to the scripts folder. So, let's open up the scripts folder. Then we can see the different tracks. So, I'm going to do it this way so you can see. So, this is the bass, this, the drums, instrumental, other, and vocals. So, if we just wanted to hear the vocals, we can just play it. Welcome to the OPT portal project, a cutting edge web application designed to transform how educational institutions and international students navigate the optional practical training application process. Every year thousands of international students face a complex document heavy opt. You wanted to hear the other tracks

[23:05]You could do that. This is the instrumental and sometimes the instrumental may not be going all the time. That's why you're not hearing it. This is why it's actually useful to open these things up in Audacity because that had vocals throughout, but it didn't have all the other tracks throughout. So, if we go to our scripts folder and we open one up. This is our base. We see we didn't have base until you know, a good minute and, you know, 25 seconds in and not a whole lot of bass overall. You also notice the bass is here. We actually had a difference between the two

[24:04]Tracks. They're similar, but this is not the same as that. If we opened up any of the other ones now, let's open up the vocals. We had vocals pretty much throughout except for at the very end. And the vocals were pretty much the same vocal on both tracks. And since the vocal dominated, you know, that's what But if we open up another track uh let's say drums, it doesn't look like we had any drums in the uh in it at all, which is kind of right listening to it. And the other instrumental will be like the the piano and stuff like that. Um and so I think it was mostly so

[24:54]This would be either instrumental or other. So it was mo mostly other. That was our little pianoy background thing. And so you can look at those but that didn't start which it didn't until maybe 14 seconds in. But if you just wanted the uh if you just wanted the vocals, you do the same thing, but you know the vocals don't do anything after there. So you just trim it again. So edit, remove special, trim. And then we have only vocals. We don't have that dead after it. And then Is that a question? Okay. So, and then you can save it again. So, just save as export

[25:53]Audio. It's going to overwrite that one. And if you open it up again, it wouldn't have that. So, then the next thing we need now that you have a script after you read your script After you read your script, um, you're going to want a mix of what you want to do. So then you can go back to So I'm going to close all this, not save it. You have your script here and then you would prompt it. So, here's our script and you go back to a tool and you would say something like um I want a mix of uh showing a tool

[27:07]And AI generated B-roll suggest uh scenes most around 3 to 5 seconds. And this would be sort of a not quite a storyboard because we don't have images yet, but this would be the basis of what images do you want in your storyboard. So, it's going to take a minute or two to do this. Any questions while it's doing its thing? No questions. Great. So, I'm going to uh open that up

[28:18]Again and put it in with the other stuff so I keep track of this stuff. And this also will allow you to uh work on it together. You can make scenes. And as I've said 20 times, it'll probably say again 20 more times. You'll almost never expect accept the first thing that comes out of there. You'll think about it and look at it. Um but what we'll do is we'll start making some scenes. So here's our first scene. Um welcome to the OPT portal. It's about so since OPT is is about international students I I would add that in the visual. I wouldn't use the university

[29:13]Logo. That's too sort of generic. Uh, and then whenever there's a screen recording here, you're just going to use something like a Snag It or any screen capture tool to just capture you actually using the tool. But I like this one, aerial view of So, there's a couple ways to do this. So, now we're going to generate some B- rolls. And the first one is this. So, I'm just going to start with that. And the way we do that is again you have access to this is with midjourney and this is under the humanitarians AI and I usually do this in the channel called YouTube and I'm just going to put this

[30:10]In a few times and it's I'm almost certainly not going to accept except um the initial ones here. I'm just going to generate a few. Another thing we can do is this. We can go to this site here. Where's the stupid zoom thing? I'll put it in here. But we could also just put this in um uh look for so this is too far away and this is why having an actual picture might be better. So I I just from Nina can comment on this, but I think you want the images to be a little bit more emotional when you use your B reel. And this doesn't say anything. You

[31:20]Should try to get a picture of Nor Eastern, right? Are you sharing it? Am I sharing? No. Yes, you are. Sorry. Go ahead. I'll share again. So that's why I would go to the Nor Eastern website, but I don't want to change my password right now this. Um, but I would say Nor Eastern. Right. Are you looking for the binder? Um, I was, but I'll just find a picture of Nor Eastern. It doesn't really matter. I'll see if I can find a link to that. Yeah, we we can share this. But you can also just go on YouTube and just find something. So, this is a little bit more

[32:13]Iconic. So, I'm going to use this picture as a base just to give me suggestions. Um, and of course zoom is getting in the way of everything. So, what I will do is I'll use I'm going to just get rid of a lot of these just to uh use a function called describe and then what it'll do is it'll take this image. It's also an AVF image, which it probably won't accept, but we'll try. Oh, it does not accept it because it's AVF. So, I'm just going to convert that image. I do this so much that I have Python scripts which convert these images. And I'm going to put it in its

[33:16]Own little file. I'm going to rename it and then I'm going to convert it to a PNG. The reason why it didn't accept it, it was an AVF image and I don't know why it can't handle that format because it's a common format, but it can't. So we converted it to a pj. And so what we'll do is we'll look at these and it probably knows anybody know the name of this Cretzman quad or something like uh the one near Richards. Yeah. This Cretman quadrangle that Krenston. Yeah. Card spin card. So, we're we're going to put that actually because it probably actually has pictures of it which label

[34:22]That again or Zoom doing its best to get in the way of everything. So, we're going to say imagine uh international students Um, I'm going to use this building is gray and white under star here. And I'm going to also add the word northeastern because it also probably knows northeastern. And I already have some default language for this. So I'm just going to cut and paste it. Where is

[35:40]Uh so to make it just beautiful photography I add this hyperrealist uh photography in the style of National Geographic. The aspect ratio is going to be 16:9 because that is the YouTube aspect ratio. This oath is for if you want to use for example a picture of yourself. So if you have old pictures of yourself and you want to redo yourself, you know, talking with Rama in ancient India, you can use or ref. Uh what we're going to have though is just a style ref. Style ref is just a sort of a general thing. So if you have an old picture of yourself and uh you want to redo it.

[36:38]Um, so it's, you know, it's basically you, but it's less like it was shot in 2025. Uh, with those level of quality cameras, you can use ORF if you, so if you want to see an example of that. Um, we'll just put it here. So for example here is a actual picture and if we use oath wherever it is that is the redone picture. the redone picture, but this is now HD, you know. Uh, and here's the original original AI redone. So, this over face, it's insane.

[37:49]It maybe 10% of the time it doesn't really keep your face but you can just as with everything in AI if it doesn't keep your face you just don't use that picture. Um but I'm going to be doing a b redoing a bunch of family photos uh with this because you know you go from okay quality to much higher quality. This is HD. Um, and that's OF. And if you wanted to, you could, uh, you know, put yourself in whatever environment you wanted to, ancient India or wherever you wanted to be. Um, but we're going to use u something called SRF. And what SRF is style ref, it it

[38:34]Doesn't mean we want to match people's faces exactly. It means we want the feel of of the image. We we don't we don't really want this image. We want the feel of the image as opposed to this where I want this weird nose. Um and it kept that weird nose. And we'll do it a few times and just start generating images. So Nick, all you do is add SRF and it'll keep the same look and feel or do you or do you need to? Sref is style ref and this is look and feel. ORF is for like putting your face, right? If you wanted to put your face in like a

[39:29]Victorian costume or whatever, got it. You would use ORF. Um, or if you just want to do something like this where you're not really you're just making a much higher quality version or changing the angle rather than having it go to the side. You you just want it that on. That's ref. Okay. So, you don't need a number after SREF. You can you can that's the weight. So both of them have numbers. A good question. So if if I wanted it to sort of look like this but not then then I can change the weight of how much it does the osf. So if I wanted something that

[40:16]Didn't keep it exactly but was sort of inspired by it but kept it kind of closed, I can add additional parameter called OW or SW which is the weight of how much does it do that? Do I want it exactly looking like that style or do I want it kind of inspired by the style? So that would be an additional parameter here. I forget the values of them. So this is where you would look things up and I believe it's called SW. I think the values are zero to a th00and where zero is like make it match exactly and so it would be an additional

[40:55]Parameter here and a th00and is like be vaguely inspired by it. Uh and this would be a weight. I'm just going to the number 100 here. And I'd have to look up whether low is matching or low is not matching. But you can just look up the references. And what that's what this other parameter is doing is just adjusting the weight. How faithful do you want to be to that picture? And I believe and we can look actually we might as well do this now in mid journey documentation. So most of these parameters like ORF, SREF, um the you can go to the documentation. Uh here's the Medjourney

[41:52]Docs. Uh where's the This is often um here. Some of these are such shitty guides that I also might do this. Um what I might do is I just might put in the actual mid Journey docs. That's a shitty one as well. And this is where Claude is great and all these are great. Uh you can put in this little prompt here. I already have too many windows and ask it. So this is where um this is where often these language models are better than the actual documentation. The actual documentation is there, but it's hard to read. But I can just say

[43:03]Explain SRF or ref SW etc. And then you can start having a conversation with it. And so what I want to know is I want to know as high is this good or bad. So what are what are the ranges of these? I think the ranges are 0 to a,000. But hopefully it will figure that out. 0 to a th00and default is 100. So if you don't do anything and that's moderate. So low is little th00and is maximum. So if I put thousand I would want to try to match that picture exactly. If if I just sort of moderately wanted

[44:13]To do it just I would maybe do it between 100 and 400 or just not change it because the default is 100. Does that make sense? Because how much do we want to keep it? Another to do is just play with it. So we can do the same thing and let's take it to a th00and and see what happens. And if the m if the documentation is correct, it's going to try to match like this doesn't even have any people in it, right? It's sort of vaguely, you know, the building looks the sameish, but that doesn't really even have the people in there, right? So, if the documentation is

[44:56]Right, and this is where I also think going to that site where they have the the actual images would be nice because they tend to have better images of Or if you want, you could literally just take a picture. You could walk outside and just start taking some pictures of of of Nor Eastern. Yeah, I put the I put uh some links in the chat for the library repository and the um brand guidelines too. There might be some images there. Yeah. And so uh you know you could just literally take a picture. The good thing about taking a picture is it will it will make it look like it was shot by a

[45:39]Much better camera. So, if I just take a picture, you know, I'm doing this a lot with a lot of old images. I'm just having them redone in mid Journey as if they were shot with a hot spot, but try to keep the picture. And it it can do that. It can um you know sometimes it makes them look too good. Um because you know if you have a family photo and you know it's shot by some random person on a flash camera, it's usually not very high quality and you can redo an image as if it were shot by a you know a professional photographer.

[46:22]Uh, but let we'll just none of these are particularly interesting. So, I'm just going to So, I'm not going to necessarily use these for, but this is again play play play. I'm going to get even closer. I'm just going to do a picture of an international student. And if you want to use your face, you can use your face just by taking any picture of your your face and using that as a no. So you can have you typing on a laptop in the whatever quad right if you wish or you can have it not completely faithfully go to your face. You can if you want to match it exactly

[47:08]You know bring the weight up to a th00and if you want it to be inspired by your face but I'm just going to go ahead and and and create an image uh just based on you know giving it an Indian name. Um uh and I'm going to use that as an image because I I right now these just don't convey any emotion to me. Um and this is not so much what we're talking about today, which is the technical aspects of building it. It just I just don't think these are visually very interesting. And you want your B-roll. So I'm just going to um international

[47:49]Student and I'm going to give her a name. Uh, and what this will do is I could say, you know, East Indian da da da da, and you can do that. But if you just give it a name, my bet is that it's going to create an Indian woman. The fact that you said student, it's likely to be doing something in the 20s. But if you wanted, you could say 27y old. uh typing on a laptop. I usually use Mac laptops because they look better uh in in a in a um in a Gothic library with a lot of books. And you're going to play a lot with the

[48:56]Images just generating it. And we're going to generate and oh, did I forget? Yeah. And I have the aspect ratio. So I did need that other stuff is I'm not using other references. So it did sort of keep this. And we'll look at at the other ones whether they were more faithful when we bring up the weight. But sort of my initial choice of a picture wasn't really that interesting of a picture to begin with. So, uh, uh, what we'll do is we'll generate a few and we're going to take these even if they're not good because we want to go to the next step showing you how to

[49:36]Make video. Again, I'll generate a few That looks nice. You know, that looks just again I didn't say East Indian. No, no, no. I just gave it a fe what I believe is a female and what I believe is an Indian name. And by doing that, you know, it get created sort of generic, you know, East Indian woman, 27y old student. And again, if you wish, you you can use your own pictures of references. So, I could put my picture in here. I wouldn't do this. It would be confused if I said this and then put in my images. The image will tend to dominate, but I could put a picture of

[50:29]Me on a laptop with the same thing around it, but that would be using the OF rather than just a generic name. So, what we are going to do now is grab one of those images. And we typically do that to the MidJourney website. And this allows you to look at the image a little closer. So, they all look sort of generic, you know, student typing in a fancy library on a Mac. Um, and so I'm just going with my instinct. My instinct is that uh you know I just like this off the ones that we have. It just feels a little bit but they all kind of are the

[51:13]Same to me. This one I like as well. But I like this. The colors are warm. Uh you know here's a student typing on a laptop. So uh if I like it, I'll tend to like it. And then we're going to download that. And the next thing and I'm going to use a different tool. You can use either tool. Stick with cloud or stick with chop here. I'm just going to show you. It doesn't really matter. Whatever tool you want to use is fine. Whatever you to you have access to. But I'm going to first uh put in the script. Get rid of all this stuff

[51:57]Here. And we're going to create something very short. We're not going to create the whole thing because the purpose of this is to we're just going to create like two B-roll scenes and then edit them and with our little voice over. But we're not going to make the whole video because the whole video is a process of just doing this and doing it again and doing it again until you piece together something. And I'll show you this in our scenes. If I generate three minutes of video, I'm probably generating a half hour of B-roll. is I'm just generating a bunch of stuff and then picking what I like. Just like

[52:35]We didn't use any of these images here. They just kind of suck. Um but these are nice, decent little images. You also don't want to necessarily go for perfection because then that will never end. Um, but these are nice solid, you know, no copyright issues because you generated the issue the image. They they I think convey here's a student typing on a laptop. Uh, and so that's what we do. Um, and so then I'm going to say use this for context and it's critiquing it. So, what I'm going to do is in the session, it's still babbling on. As I end the session, I will upload images.

[53:49]You will describe and suggest two shots. each with image and I'm going to just capitalize this parenthesis and a suggested camera movement. uh for that works with a script. So then I'll do that for this one here.

[55:18]And we'll do that. Actually, I want So, I'm going to just correct it. in the session. Reason I say in the session is I don't want to keep on typing and typing and typing and telling it to do the same thing again and again. So now this is better. Shot one, shot two. So shot one describing the scene. Shot one is also talking about the camera. And then we're going to do this for for each one. So then we go to runway which we'll also give you access to and this is for B-roll you whenever you do stuff. So what we do is we generate

[56:08]Video. I'm just going to call this tutorial and we upload the image. You can also Photoshop it or do whatever you want before uploading it if you feel there's a need for that. But generally there's not. So we're gonna do shot one and we are going to do shot two. And this will take about five minutes to create these two shots. And this is unlimited as we can see here. So the limit is in you can't do too many. So I can do up to two at a time and I have to wait for these two to finish before I can cue two more. But this will

[57:02]Take about 5 minutes. So let's generate a second shot from the script here. Um, I'm just going to create this and I'm going to put in some words. Go back to my journey. Get rid of that. That rid of that. Uh, I'm just going to just start pasting some of these things in just to get a second shot. Most of the time is really thinking about these things. Um, but because this is a tutorial, I don't want to spend I'm going to keep the same uh I'm just going to remove the library and the typing. I'm just doing this to get a second shot

[58:13]Because we're going to need two shots just to show you how to edit. I'm not going to worry too much about this being the perfect shot. But this is most of the work. Most of the work is your storyboard. Uh thinking about the shots you want. Um doing that. And for a lot of what you do with a real app, if you do a minute video, probably 30 seconds of that video is you just showing the app. Probably half of your your which is actually a little easier to do the showing a tool videos. Um because this is a vague prompt, uh we're getting a lot more diversity.

[58:52]And you can see whenever you get something where the four are very different and um the library looks like it's still in there somewhere. Um but this this I like this one here. Shows some confusion. And this one here I I like actually I like this one. It's kind of a good shot. It's showing somebody being overwhelmed and just bureaucracy and paperwork. A recent graduate remarkably good for such a vague prompt. So to me, Nina can comment on this, but this is sort of conveying recent graduate being overwhelmed with all this bureaucracy and paper stuff. So I'm going go ahead and use that. Um, so we're going to do the same

[59:49]Thing. We are going to uh download that and we are going to um go here to just two shots. And again the structure of these shots are what's in the scene and what is the camera movement. You can also add effects. We'll do that in a second as well. So, we're going to upload the second one and we're going to have a third shot with some effects that we add on top of it. So, let's look at our first one. Don't know why her laptop is shaking that much. That looks a little better. So, I'm going to favorite it and

[1:01:01]Download that one. And we'll wait for the second shot and I'll start the third shot uh once the these two are done. But again, you're going to mix this probably somebody typing and then you're going to show the actual screenshot of the app being used. Couple minutes and we will get There we go. I'm also going to add a little effect. I'm going to say overlay. Uh, I'm just going to say, uh, dollar signs. So, now I'm going to add an effect that isn't in the original image. Swirling around dollar signs and question marks. So I just add over overlay dollar signs and question

[1:02:12]Marks swirling in front of her eyes. So this is not something that is in the original image. Uh but what we're asking is we're asking the the um system to generate that. So we'll generate a couple of those. We'll take a look at these. Okay. And again, you're going to play with these things. I don't know what the smoke is coming from, but I like that one a little bit better, nevertheless. And then what we do is we open up some editing tool. You can download Premiere for free or whatever tool you want. And we are going to remove these here. So, we're going to import our

[1:03:17]Media. So, our media consist of these two things here and those also include other shots. And I'm going to zoom in a little bit so we can see this a little bit better. And our our voice over, which is here. I don't know why we have Oh, I know. Wait, we have two. We have one that's a little smaller because we Oh, this is just the vocals. So we can see how many we need. We generated two shots. It looks like we need maybe 30 shots uh to complete a two-minute video, but two minute is kind of a long video. We probably want to, you know, keep it 2

[1:04:18]Minutes or under or something like this. Uh but when then we just welcome to the OPT portal project a cutting edge web application designed to transform how educ. And so what I tend to do at this point and this will be the actual work is welcome to when I hear that I I I probably want a scene that indicates like welcome. So, the script didn't really create this, but we'll go ahead and create that now. But, so here, welcome to I don't I like this image, but I don't think this image belongs in the first 3 seconds of this video. I think welcome to the the the OPT

[1:05:06]Portal project could either be showing the actual project or if we want to use a visual metaphor something more like a more welcoming gesture like a handshake or something like that. Uh and then this is the process. The process is listening to that. So the way I do it welcome to the opt portal project. So in my mind and people you know can have different views on this. I like this image but I don't like this image there. I don't like it at the beginning at the welcome but let's just continue on a cutting edge web application. Cutting edge application there. I want some sort of tech thing, some sort of

[1:05:49]High-tech image designed to transform how educational institutions and inter here educational institutions, things like this could work. So this one could be here and not necessarily the beginning. Um because this is so these two could work there and this is the process. So, we're we're not going to make the whole video there because the whole video will probably take about five hours to do because what you're going to do is you're going to look, okay, where do I want B-roll and where do I want the actual image of the actual app? And then you might even look at your your voice over and decide, I want

[1:06:34]To change my voice over, in which case let's actually change the voice over to match this. So if we were to change the voice over to match that, what we would do is we would go back to suno and it would be something like overwhelmed by uh where are we here? Um how many things So, I guess I want this changed, but we're going to generate that. I'm just going to generate an 11 laps. Where's my

[1:07:44]11? And let's use an Indian voice. Actually, I'll use Emmy. remains fine. Every year thousands of international students face a complex document actually because it's an Indian face. I think I might pick an Indian voice. We can search the language and accent. So the language is English. accent is Indian. Just don't give up trying to do what you really want to do. Where there is love and inspiration, I don't think you can go wrong. I don't believe that. But actually, let's let's use this voice. Wherever you are, curled up in bed, gazing at the stars, or missing someone special, stay with me. Every year,

[1:08:55]Thousands of international students face a complex document heavy OPT application process. Universities struggle with manual tracking systems, fragmented communication, and administrative bottlenecks. What if there was a better way? We're going to use that. And so we're going to change the voice a bit. And you can do your voice not necessarily in one full thing. You can do it in steps if you wish. Don't have to generate the full voiceover. But we're going to import that one. And I think these two scenes actually might work for for this little bit of stuff. I'm going to zoom in to get better control. Sure. I line up exactly. And then I'm going to create a

[1:10:06]Transition here. I'm just going to transitions paid also fade there. Every year thousands of international students face a complex documentheavy OPT application process. So in my opinion that voice over matches that image much better. Nina is the expert on that. Um but that's sort of the process of doing things. You want your visual to match what's being said and with everything in and so let's just play what we have so far. Every year thousands of international students face a complex document heavy OPT application process. Universities struggle with manual tracking. It end up fading through black. I don't know why did I add the wrong fade. I'm just going to

[1:11:15]You can also see where the voice is. We generally want to change scenes on empty spots. So, we're going to do that. Every year, thousands of international students face a complex document heavy OPT application process. universities struggle with manual tracking systems, fragmented communication, and just keep on doing that. And then at this point around here, I would sort of add a picture of the actual and begin to to show the actual product, the OPD portal. And so this is the process. Um, and we just keep on doing it and doing it until you're comfortable with it. You may only like this much and then generate another scene there and then go

[1:12:08]Here and generate something and there and generate just a lot of back and forth like everything with AI. So we changed our voice over because and I actually like this intro a lot better than our original one because it clearly states the problem. There's this massive, you know, red tape bureaucracy thing. Can can we make that easy? Um, so every year thousands of international students face a complex document heavy OPT application. Let's see if I have one that's more completely done. I may not because I tend to move because these things use a lot of space. I tend to move them. I don't think this but I think I

[1:12:57]Have the video not this but the actual let's import that import it. So after doing this for about four or five hours you end up with um something that looks more like this. So this is basically just scene after scene after scene after scene after scene. This is actually a long one. It's seven minutes. Um but it just unchained. So this theme is just you know this is about uh pride month. So, it's just sort of pride back then when that started with the Stonewall riots and pride now, which is much more of a celebration, but back then it was people getting beaten and and put in jail and all that kind of

[1:14:08]Thing. And it just goes back and forth like that. There's then now, then now, then now, then now. Uh, and you just do it again and again and again until you think you're done. And you're never fully done. Uh, because you'd always look at what you do and say, "Well, I could have changed this, but at some point you're just going to, you know, have to be done." So, that's the process. Uh, any questions on on the process? I have tons of questions. If I can jump in. So this is very a very different workflow from the traditional um you know way of building a video like

[1:14:50]This where you start with well you start with a script right um for your timing and then you storyboard everything out you have it all laid out before you start you either shooting or generating. Do you tend to sort of build as you go or do you That's an interesting concept as you're right. The difference is the reason why you did that is you just to those scenes, you know, any of those scenes would have cost you thousands of dollars. Yeah. Money. You have to plan it and pre-plan it and pre pre-plan it to make sure if you're going to shoot that you're going to use that, right? the the

[1:15:28]AI processes just generate something. Do I like it? And a little accidental like that that uh image of the the person getting buried, the student buried in all the paperwork. That was just a random thing that came out of it. But I kind of liked that. So I take and then I went back and changed the voice over to better match that image. But I I think you're right in this is and this is something we should write about and talk about is the process I think is very different because it's so easy to generate you know the voice so easy to generate uh I don't like that voice it

[1:16:07]Should have been an Indian voice not a you know a New Zealand voice right but you can't do that in the old style because you had to hire a voice actor and once you hire the voice actor that's your voice you can't go back and change that voice for that particular shot. Here you can. Here it's just literally, you know, going back and forth, going back and forth, going back and forth. And a lot of it's really your eye and your feel of, you know, being able to quickly generate stuff. So, for example, with you, you're learning this stuff, but it may take you a while to generate any of

[1:16:44]Those voiceovers or any of those images, but as you get better at it, you'll you'll be able to generate sort of what's in your head, very quickly, right? Um, now it may be, well, how do I generate an image of, you know, a student on a map Mac laptop or whatever you're thinking. But as you get better at that, as you learn to do that more quickly, I think your process will be different from the regular storyboard process because even in the regular storyboard process, I'm bet I wish I could have redone that better. Oh yeah, absolutely. You know, we're $40,000 into this. This is committed. We're we're doing this and

[1:17:29]Getting it out the door. Um whereas with this kind of process it's it's a matter of you know playing with it and trying playing playing trying trying playing playing trying trying playing playing trying trying playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing trying trying trying playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing playing because any of these things and that's really the the difference with sort of sort of garbage AI and good AI is the garbage AI is the people who take the first thing they get

[1:17:50]Out of chat GPT and claude and always use the first thing they get out first image first script first voiceover first this first that and And but if you don't do that, if you say, "Hey, I could may, hey, can we rework this? Maybe we want a, you know, this in the scene. Maybe we want that in the scene and play with it, then you're going to get a much better quality product." So you're you you don't have to plan out, okay, I'm going to go from a tight shot to a medium shot to, you know, you can you can change it as you go along. Do you is that your

[1:18:25]Process is just build as you go? just really the process you saw is fascinating. Never done it that way. I didn't think about it being so different from a regular Oh, it's totally different. I mean, I I don't even know how I'm going to how how I'm going to approach it because it's so different from what I from my process of spending lots of money, hundreds of thousands of dollars on a, you know, on a commercial. Um, you have to plan everything out, including, you know, your whether it's tight versus medium versus, you know, you go down to the nth degree. You do a shooting board and not just a

[1:19:04]Storyboard. You go and shoot, you know, it's like second by second, frame by frame. So, because you're spending lots of money, you know, you're really playing with things and seeing what the things can do. like you know the the stuff with playing with the old family photos you know just hey can it do this? So then what I started doing is just putting pictures. I took a normal photo of myself and then put myself in scenes of me in the Rammanana. And whether that works or not, you just look at it. And again, if you put yourself in the Ramayana, that can open you up to other criticism

[1:19:43]About, you know, you know, not, you know, cultural stuff and other stuff. But it's amazing what these things can do. It's crazy what they can do in terms of you could star in your own movie and it's they'll turn your face. I mean this this is it might not seem like much but to be able to take a photo and then now that it's a photo I can make a video from it. So to take a photo like this one and then turn it even a little bit or you know put it in a different outfit um uh here just wasn't possible two months ago. And so, you know, you have you can

[1:20:48]Have your own opinions on how faithful this is. That looks pretty damn close. Yeah, pretty amazing. And so and but that switch of you know a little bit and of course it's upscaled and and higher quality as well which is nice if you just want to restore uh restore old folders that that's stunning. And then you can do things. It doesn't really work for this, but uh like on on the mid journey here, if you wanted to have fun, and this is a bit fun that you can have for your parents. I don't know how people are familiar with this particular movie. If I can find

[1:21:42]It. Uh but a movie called Home Alone. So that you were and you can see the original photo there, right? And you can see, you know, you as you know, home alone. Yeah. Like that. And so it's it's insane what you can do with these tools. And I would just to make your parents happy, just send them a picture of you, you know, in something that is meaningful to them for some reason. I don't know if home alone is meaningful for them. Uh, but it's, you know, it's insane what you can do with these tools. I I have one last question before um before I monopolize this thing. So, I

[1:22:37]Think the one thing we want to do is I like the outro that you created for humanitarians and I think we want I mean in terms of anything that goes on the humanitarians YouTube, we should probably have the same outro. So, let me put that before I forget. Let me just put that on. So, the outro is here. Um I think it's here. No, I have to remember where I put it. This is why it's also good to put things on. I'm trying to figure out where these things are. Yeah, here they are. And this is another reason to put things on. So, I'm going to put this on. I'm going

[1:23:22]To call this outro. And I'm going to put it on the GitHub right now. And put it on YouTube under YouTube. Uh I'm going to actually call it art. I'm going to put it under art so it'll be easier to find. So all we do, let me show you. We'll we'll add an outro right now to the current video. Let me just upload it here. So to add an outro, all we do is this. We uh um go here. So we wanted to add an outro this

[1:24:34]Where were where the ending of it. We go about there. Uh we import since this is dark uh generally dark. I'm going to import the white version. If this were very light I would import the dark version of it. Uh, but we just go to our art and we go to outro. So, we we do the high-res version of it and we put it there and then we add an effect to it. And so you can see that if I wanted to end exactly where this ended, I would do that. I would move this so these align. And then I'm just going to add a

[1:25:27]Transition to it. So I'm going to add not a fade. This time I'm going to add a dissolve. And at the end I'm going to add a fade, but I'm going to do a fade through black. Then I'm going to zoom in a little bit. And I'm going to make this one a little longer. It's going to take longer to fade in. And so you can choose any effect that you want. And then we also want a title done in the same way. So the title you would create in systems fragmented communication and administrative bottlenecks. So the this has to fade to black as well. If if this were actually the end,

[1:26:21]It would fade to black. So then this is the way we're going to end everything systems, fragmented communication, and administrative bottlenecks. What if there was a this is going on? So if that were the ending, that's how you create the ending. If you want a title. Uh, did we agree on Garammon? Yes. If we want a title again, Illustrator, this is a new thing, so I don't see Illustrator, but I think I installed it. So, we we just create what I need to give you is the specs. Off the top of my head, I don't know what the specs are going to be. So, I'm going to actually

[1:27:17]Import the one I have. I'm going to open and then delete it. The reason I'm doing this is not because I need this image. I need to know how big this image is. And so, I'm just going to import this mainly for its size. Uh, and I'm going to call this intro I'll also uh and I'm going to call this intro. And this is just going to be in a illustrator file. So I actually don't need this image anymore. So, I deleted it, but I I'll just give it some title. So, I'm just going to go to the type tool and it needs to be much bigger than

[1:28:17]That. And it needs to be gammon. And there's so many Gmons, we should probably decide on a type of Gmont. But I'm going to go with the semi-bulb maybe subhead. Subhead is nice and bigger. And you can make it as big as you want. So even though I can only take it to 72, I can just type in whatever I want. My guess if I wanted it the full thing, it's going to be around there. Little too big a guess. It looks zoom. No, perfect guess. That'd be a little too big. Take it down a bit. Of course, this needs to be white. I'll just keep it black for the

[1:29:26]Hell of it. what that looks like. So, we save it. Sport is a ping. Make sure it's transparent. So, I I think it by default it'll be transparent, but I'll just export it a little bit more. You're you're talking about the background. Yeah. Transparent. Okay. Yeah. So, it's transparent here. Uh and I'm just going to put that in the same Where where is this? I think it's in the downloads, but we'll see. And this will be the the the beginning. So, we go to the very beginning and the title will be something like OPT portal. It wouldn't be Lauram Ipsum. Uh, and we do the same thing. We

[1:30:13]Just uh import I think it's here. What did I call it? intro or maybe it is under the scripts. I don't know. Is it under outro? Yeah, there it is. Intro the bottom. And you add it uh intro. And then you do the same exact thing. You just add some effect. I'll add a different effect, different transition to it. I don't know what 90% of these transitions are. And a lot of them are really funky. I tend to just use fades and dissolves. Yeah, I do too. Uh but we'll just do one for the hell of it. And then we need to to I can just fade

[1:31:03]It out. I'll just dissolve it out. And there we go. Every year thousands of international students face a complex document heavy op. And that's probably a little long. So, I'm going to shorten that. You should be able to read it. So, basically, um, do you know any rules for this? Like read times three or something. Takes you three seconds to Yeah. Every year, thousands of international students face a complex. That seems long enough for a title for me, but you should probably write down some rules, Nina, about title length. Got it. And the title length is probably affected by how long the title is. So, it's a little longer. It might

[1:32:00]Be a, you know, it's a longer title. It might be a little longer that you keep it on screen. So, I think that's the only thing we want consistent uh on all our videos is the intro and the outro, the use of the type face, just so that there's some something that holds all these videos together. Yeah. and I'll I'll upload the the AI. So, you have the you can basically just open up an Illustrator and replace the Lauram Epsom text with whatever your text is and it'll keep the the the same size and it'll keep the uh um same type face assuming you have the type face

[1:32:42]Installed. Uh, I thought we put this in things in our Oh, yeah. It's under outro. Does everybody have the Adobe Suite? I know it's it's free to Nor Eastern. So, you just you can type in Northeastern uh Adobe Suite Creative Cloud and it'll show you how to install it. But it's easy to install and I think you are able to put it on three different computers. Yeah. So this intro white I'll also well you can change it also just for the hell of it just create an intro black as well. Um and put those there. Uh yeah, but um yeah, so just uh replace intra white

[1:33:42]Or intra black with whatever your actual title is and just have a some transition in and some transition out and Nina will write down some rules which you'll put on that page about the exact name of the gmon um you know gamon semi bold. Yep. and and um the the 300 point and um uh if it's you know this many words versus that many seconds kind of you know heruristics. Thanks for listening to this recording of Studio Sessions with Bear, a live humanitarians AI training led by Professor Nick Bear Brown recorded on May 17th, 2025. We hope it gave you practical insights into using AI for educational

[1:34:34]Video creation. If you found this helpful, don't forget to like, subscribe, and hit the bell icon to stay updated on future sessions. Share it with a friend or colleague working in AI, education, or creative tech because learning is better when we do it together.