Latest / And Now For Something Completely Machinima / S6 E247 RUS-T - Can an AI Film Tell a Story? (Oct 2026)
Transcript
- Phil Rice 01:00 Hello and welcome to and now for something completely machinima, the podcast about machinima virtual production, and as we will see today, other stuff. So this is a I think in some ways unique pick for us. It's not the first time that we've covered a film that involves the use of AI technology, but the particular way that this was made kind of highlights some some new tools, and as as well has an outcome that is not what you would normally, I not what you're typically seeing from these tools so far. And this is Tracy's pick for this month. So, Tracy, why don't you tell me about it? Tracy Harwood 01:43 Yeah, absolutely. Well, this has only just come out, actually, 31st of August. The film's called Rusty, AI sci-fi film created with Reillusion AI Studio and Seed Dance 2.0, and it's by John Martin, who's obviously a director of Reillusion and the creator of the AI studio, so he's testing out the tool as well as creating his own piece of work. Now, yeah, have to say, you know, given this only came out as we're reviewing this two weeks ago, 13 days ago, in fact, this has got over 750,000 views, which is, I think, an extraordinary view count for you know an AI generated what what appears to be an AI-generated film, but I want to break that down a bit because it isn't that. There's a bit more to it than that. And what I think you've got here is, you know, John creating a a work both in a personal capacity and as a as a workflow test for a new toolset, so I want to start with the film, and then I'll talk about the the toolset and the way he talks about it in his presentation of how it's worked. So it's and by the way, this is you know I'm obviously going to do a bit of a spoiler here, so you know if you want to watch the film first before you listen to what I've got to say about it, by all means do that. the The link will be in the in the description of our episode here. It'll also be on the show notes. So by all means, do that and then come back and have a listen. So this is a science fiction short about a robot on Mars, and I think what's interesting to me about it is how its visual storytelling connects youthfulness and abandonment and care in in in the story that's being told. The opening gives us an enormous orange landscape and a robot, presumably rusty, and this robot is attending to equipment in the middle of it. It's a really good introduction. We meet the robot through the work that he's doing. This sort of this this robot has this kind of weathered body, which really tells us something before the story needs to explain, you know, what it is that it's doing. You know, it's scratched. You know, it's got exposed mechanisms. It's got discolored metal, and it makes it really look like the robot has been used for for quite a long time. I do like the contrast between the worn exterior and those kind of bright blue eyes that you see, the eyes really give it a very human and it and an immediate kind of you know place to to look for responses. It is really humanized. It's an effective character design, although I think it also has a very you know familiar appeal to it. You know what what's being done here is the machine has these very you know it's very sympathetic way of presenting a machine because its its face is very legible to us in in human terms. In fact, what's also striking is just how humanoid the robot is as a as a machine. Because you know, there's at times when it's when it's moving, the skin actually isn't really very robotic and machine-like. It's more like it's got you know the skin is its clothing, and it's it's it's it's got a sort of a fundamental aspect to it, which is much more than when than human. It it moves very, you know, it moves the skin moves the face moves everything. It moves like it was actually human, not not machine at all. The outpost adds to that impression. I think you've got cramped interiors and a utilitarian equipment that suggests this kind of an inhabited workspace, and you've got that set against an open desert, and these these rooms kind of promise shelter. But later, when that promise fails, the setting itself becomes part of a betrayal, which is conveyed quite well in this. I think the reservation is that the world feels very familiar: orange Mars, battered machinery, isolated worker, expressive robot-these are kind of recognizable science fiction tropes that are that are being drawn on. Although, actually, what happens here is that that helps to convey what this film is about really quickly. Tracy Harwood 06:57 But it means that actually the character's choice choices have to provide the originality in the story because the production design alone can't carry that burden. So, what you've got there is a very particularly sharp visual moment in the in the mission documentation, we see instructions to abandon the search for life, evacuate, and return to headquarters. And then another kind of close up tells us that Rusty will be destroyed with the outpost. And the wording on that message that you see makes that brutally clear. The crew can leave. The machine is part of the property being disposed of. But already, what we've seen is that machine isn't isn't a machine really. It's a human. So that's a strong message in in what's being portrayed. It's I think the the strongest ethical idea in what's portrayed actually what you've got is the responsibility of you know do robot creators you know what what are they what are what responsibility do they have for the robot itself You know, when they've relied on the the labor of something, do they just then discard it? That's a question that I think pops up a lot, not just in science fiction, but in the way that these things are used, even as physical entities in in today's world of work. Actually, it's quite an interesting. Situation to find yourself in. You know what we've got here is a story that's telling the story of you know the the evidence of of service, but service has not brought brought it protection, even without you know making the the claim about real AI consciousness, this this kind of works as a story about the treatment of these kind of dependent workers. The written notice is a really efficient exposition. Although I would have liked the conflict to be less dependent on spelling everything out. The document kind of communicates the threats, but people behind that decision remain harder to understand, and a more complicated relationship with the human commander could make that abandonment more painful and and less schematic. I thought the escape imagery moves us towards familiar action film territory. The crew leave. The robot is upset and seeks out some kind of revenge. But my interest returns most strongly when Rusty reaches a lake as he's kind of, you know, moving across this sort of landscape. And I think what you've got. Is an interesting presentation of kind of scale because you you see catastrophe in in the overarching presentation, sort of reduced to a hand and you know a finger touching the surface of a lake. That's a nice touch, actually. It's a nice way of playing with scale, and then you've got this little creature that's living in the lake. Emerge from it, and it's got really large eyes, and it's really vulnerable in appearance, and it's you know it's it's clearly you know invoking your sense of wanting to to care for it, and I think that's that's quite a strong draw in what you're looking at. Rusty kind of touches its head and the final image of of them sort of alone in the landscape together, and you know, I read that as kind of a new relationship being formed and giving this robot purpose beyond what the institution has assigned it through its work. And it's kind of warm, but it's also a very kind of reassuring resolution. That I mean, this creature is clearly engineered to be lovable, I would find the encounter perhaps more challenging if the life the robot discovered were strange or difficult or resistant to affection. And I guess it kind of asks you to think about, you know, whether you would you would still care for something if it wasn't cute, so I think the ending could have been pushed a little further than it actually was. There are a few kind of continuity issues for me, things like breathing. Does a robot actually breathe or blow on things? I don't know. That seemed a little strange. The way that you know that bodily function was kind of portrayed, although of course what what you're seeing does actually immediately build empathy because it's one of the indicators of humanness. And then for such a dusty planet, there were no footprints, which I found broke me a little bit. Tracy Harwood 12:22 The spaceship is different in pretty much every shot that you look at, and when it took off, there was no kind of support structure in the shot. But you know, it's every shot that you then see it's in. There's this support structure. The orange planet for a scene or two miraculously turns into Monument Valley, which I love, by the way. And then the the water representation is a little odd. Kind of makes you question why did they abandon the planet in the first place. So that was you know there's a few sort of strange things in in the way that it's portrayed. The creature is also quite cartoonish. It actually reminded me a little bit of the lizard monkey thing in the '98 Lost in Space film, the Blarp character. And then finally, the sound design. There's nothing in the description to give us a clue how this was actually created, the soundscape was actually created, whether it was live acted or generated. So it's a bit difficult, really, to comment specifically beyond its face value. I'm guessing because it's not credited, it's generated, which is something that Seed Dance is capable of now. Obviously, you have music and foley mixed here. I'm not sure that that aspect, you know, or or whether that's you know, I'm not I'm not too sure whether it's been layered in after the generation of images and sound. I'm not I'm not really sure exactly how that bit was made, and there's no real clue as to you know what what was done there. There's nothing in the in the in the notes that John's left for us. So I'd be interested to hear a little bit more about that. I mean, generally, I think it's it's a solidish story. I think the music gives it tempo and actually lifts it a little bit. Now let's turn for me to the the creative process because I think this is the interesting bit of it. Now, it lists character creator, AI studio, nano, banana, and seed dance. Now, the director statement reproduced alongside the online video gives more specific account. The robot was developed in Character Creator and then adapted into an AI actor. The generated frames and multiple views of the environments were created with Nano Banana, and those images were used as start frames and additional references. Or seed dance 2.0, the spaceship came from meshy with multi-view imagery used to keep it recognizable. Now ChatGPT, as I understand it, helped refine action prompts, and Premiere Pro handled editing and post-production. So that tells us where the control came from. The robot began as a kind of a reusable 3d design. Its visual identity was then carried through reference images into generated shots. The film is assembled from several kinds of authored and generated materials. I guess we should also be precise about what the statement does not establish. It does not say that every shot was animated in an iClone, and then passed into seed dance as motion reference. John's account specifically emphasizes prepared images and prompted action. So I would describe Rusty as a kind of a hybrid 3d character and image to video production, without assuming a fully animated 3d scene existed underneath each of the shots. Now, Reillusion AI Studio is the production workspace. Seed Dance is one of the generative models available through it. AI Studio can organize actor, image, video, and audio references, and associate those references with instructions for particular shots. Its reference workflow, according to the manual, lets a creator describe who appears, what they do, which environment to use, and how dialog references should apply. There is also a more explicitly 3d guided route. The you you arrange and animate a scene in the 3d viewpoint, capture its video, and supply a styled first frame image plus actor references. The capture animation gives the model a visual guide to movement and composition, and reallusion even describes using simplified dummy characters, so the motion can guide the output while actor references provide the appearance. Now, for machinima makers, that is a significant possibility. Skills in staging, camera placement, performance can become inputs to generation, so I think Damien would be particularly interested in hearing a little bit about how that might work. You can communicate an action through a scene you've made as well as through words. That's probably the bit maybe he's less interested in. Tracy Harwood 17:36 However, the reference video is a is a guidance model. It should not be treated as as proof that the generated output preserves every joint position, object, relationship, or frame of timing. I guess we need to judge the result against the reference. The original 3d scene remains editable. The generated clip is a is a kind of new piece of footage that comes out of it. I think other documented functions include generation between start and end images and tools for dialog, image editing, and enhancements. These really are capabilities of the of the wider workspace, not necessarily evidence that John has used each of these in in Rusty, there is a version detail worth keeping straight. Actually, AI Studio 1.1 now runs as a standalone application, and Reilusion's current pages include C Dance 2.5. Those later features should not be projected backwards onto a film identified as using Seed Dance 2.0. I don't know how significant that might ultimately prove to be, but it's worth noting that that's not what you're going to get now. Seed Dance 2.0 is ByteDance's audio video generation model. Its launch documentation actually describes inputs from text, images, video, and audio, including up to nine images, three video clips, and three audio clips. So those can guide appearance, composition, motion, camera movement, and sound. It supports multi-shot audiovisual generation of up to 15 seconds, alongside video editing and extension. What you've got here is how long is this? 1010 minutes, 12 minutes or so? Phil Rice 19:34 Eight and eight and a half. Tracy Harwood 19:35 Eight and a half minutes. So it's you know there's a lot more going on here than just an AI generated piece of content. Now, AI Studios reference workflow. You can you can in it assign those inputs to the elements you describe in the prompt, and the reference limits also depend on the workflow. I think the 3d guided route reserves it. Inputs for the the first frame and the captured video, and that's why different interface pages can show different reference counts without necessarily contradicting the model's overall limits. I think for Rusty, the documented image use is is kind of it. I describe it as image led video generation, I think what you've got is a an you know an established shot, which I think is then used within seed dance to generate action from from that image and then the supporting references, and I think that's quite different from asking the model to invent an entire finished film from a single sentence. There's a lot more craft and workflow that's gone into this than ever that would suggest. I don't know for sure how the sound was was was created as I said. I don't know how the the musical effects were produced. I'm not sure if we can attribute any of that soundtrack actually to the AI side of things. Although one thing that I did notice is that lip sync can be part of the seed dance workflow. I guess what what's not clear, and again, this may be something for John to comment on, is if you're doing, you know, retakes, how that workflow of lip sync carries across the retakes. I'm not really too sure how how that even worked, or whether it did, and if if it didn't, how you then manage that kind of process. I think ByteDance also acknowledges quite a lot of problems with detail stability, and you can clearly see that as a problem in some of the scenes here, too. You've you've got issues of multi-subject consistency, accuracy, complex editing, and occasional audio distortion. I didn't hear a lot of audio distortion, but I definitely saw some of the the the models wobbling a bit in in places. I have to say, I think it's an impressive demonstration. Notwithstanding that this, you know, there's a lot of limitations that are being worked with here. I I think you can kind of you can you can see the limitations evidently in it, but I guess my practical question is going to be about repeatability. If the filmmaker asks for a smaller reaction, or or a different gesture, or a camera move two two seconds later, so to speak, is the tool capable of making that revision whilst preserving, you know, everything else that's being worked on. I'm not sure. Not sure that comes through strongly enough. Maybe, maybe that's something to, you know, you know, ask ask John about and ask him to comment more on. So yeah, I mean, I've got, like I said, I've got comments on on both the story, which there's room for for improvement for sure. But it's not a bad starting position, and you know, also, you know this is clearly a workflow test. I guess a question you might sort of ask is: Can you do those things simultaneously? Can you can you create a story and have it work as a story? And can you test out as a workflow at the same time, and and show it as a workflow test. And I, my gut feeling is that John's trying to show both of those things simultaneously, and I'm not sure it's fully succeeding on either level. If I'm brutally honest about that, I think it probably works better. Well, I don't know. It's do. It's clearly doing both things. I don't know what you think, Phil, but you know that's my, I suppose my overall take on it. Nice piece of work. It raises questions. Over to you. Phil Rice 24:38 Yeah, there's a there's a series that was on HBO years ago. I've referenced it on this show once or twice, I think, but it's called Raised by Wolves. It was Ridley Scott produced science fiction set in the future, and it features these. I want to say biomechanical androids that have this-they have flesh components, so to speak, like they and very expressive faces, and they move rather fluidly, but they're they're they're synthetic 100% I kind of feel like maybe, assuming that that the kind of stretchiness and stuff that we see, not just in his face, but sometimes in the body movements, maybe that's that's what he is intended to be to inspire, because that the the characters in that raised by wolves very much the same way. There's pretty much a full range of emotions that they're capable of, and we see a full range of emotions from Rusty here. I mean, really, Tracy Harwood 25:57 we did all the Phil Rice 25:57 major food groups are covered. You know, there's times when he seems pleased. The times when he's angry, you can tell that he feels betrayed. He's determined. He's surprised. He's afraid. It's pretty impressive range of emotion shown on the character. And I kind of, I just, I think, just made a decision when I saw how fluid, particularly in the face. You know, he's got this what looks like a metallic part of his silvery part of his face here, and yet the eyebrows conferrow there. So I thought, okay, so this is some other material, other than just you know steel. It's some kind of a synthetic material, which means this is very distant future, right? This isn't something set in 2027 or something. So I I didn't I didn't find that I was I think once I just kind of settled on that, then that didn't bug me a whole lot. I do like the fact that that you pointed out, and it's it's illustrated very early on and kept nicely consistent throughout. Is everything is worn and dusty exactly as you would expect it to be on a desert landscape like Mars, you know. So everything has that dust and looks worn, and and it's it's great. Even the equipment that he interacts with, and the outside of the habitat, and nothing looks shiny and new, which would which would have been a real major immersion breaker. If you have this, you know, dirty landscape and this super clean robot, it just wouldn't make any sense. So, I think that was deliberate, and and I think it was was was effective. I found the story great, and actually, I have what may be an explanation for the the question that you was raised for you about the water-that if there's this water source there, why did they leave? My assumption from watching it unfold was: well, they didn't know that they've abandoned the planet because they've determined there's nothing of value there, and kind of by happenstance, as he's disposing of those explosives. It opens up some unknown fissure in the ground, and that's when that lake appears. Now, the the irony of that doesn't like the how that ties into the ending. It doesn't quite land as a punchline, if you will, because it's like, okay, so now what? There is life on Mars, it turns out, and it's aquatic or or amphibious, and there's there's water on Mars. But so what? What is Rusty going to do now? Is he going back to call home and tell the humans? It's going to rust, right, right. So that part just-it's just a little bit a sense of this was a major reveal. I think it was executed wonderfully, like that. It was totally by surprise, Tracy Harwood 29:19 yeah, Phil Rice 29:19 to me watching it. That just, oh wow! So there is water. I get it now. They shouldn't have left. Humans were hasty. They betrayed him, but also the whole premise for them leaving was they just weren't patient enough to keep looking. That's kind of the vibe. But it's like okay, so now what? I mean, he doesn't benefit from water at all, from what I can tell. I guess he could clean up a little bit, but anyway. So, but overall, I mean, the the the story, the the way that it played out, the arc of it, it would have been interesting to see. This whole story told with a android that did not convey as much emotion to see how well, like if you only could do it by body motion, if if the face didn't give you a whole lot, I think it could still work, right? But because they employed that, because John employed that dynamic facial expression consistently throughout it, it's one of those things where okay, very early on he's showing curiosity, his eyebrows are raised as he's looking at the thing. That because it was consistent, it it it just kind of felt right for the story. It's like okay, this is a this is science fiction, so you've got some liberties. This is a a rule that has been decided on for this story, and and it stayed consistent with it, and that's good, you know. So, um, you can tell this was made with AI, but it's certainly not slop. You know, it it it it it does have a feeling. It actually conveys a feeling that there was more control than it turns out that there actually was, you know, I was I was genuinely surprised to reading through the notes and then hearing you comment on it too to discover that a lot of this was and maybe all of it was image guided AI generated video. I when I watched this before reading that, my assumption was okay. So we're seeing now what it's like to guide the animation with something like iClone, and then run it through AI, and it gives it this extra sheen, kind of the the whatever that technology is that NVIDIA is rolling into certain video games, it's kind of controversial. The DLSS or something like that, that basically it's a render thing that's happening in real time to make these maybe more primitive characters and environments look shinier and and more real, more photoreal, right? So I was when I watched it, I thought that's what we were seeing. But he doesn't list iClone as a tool at all, which iClone is where any of the keyframed animations, the real control over the animation would happen. So on one hand, as an iClone guy, that's a little disappointing because I was, and again, I have no real plans to to make AI a big part of my production thing. But that would have been interesting to me. Of okay, as a as a almost a style transfer to my own animation, that's interesting, you know, because it's not reworking my whole workflow. I don't know how that would work because it's when you're doing image-based video generation, you're paying credits on like a per clip basis. But if you were to try and do iClone through AI and end up with an AI rendered version of your iClone animation. You would have to do it every single frame as an image, and then they would piece together. And I would think that the expense of that would be extraordinary. I think wouldn't you definitely Tracy Harwood 33:38 a beginning and an end image? Phil Rice 33:40 Well, not as not if you want complete control over how he touches his hose, or which you know the the kind of precision that you can get in an iClone environment. You're kind of abandoning that to let's hope that the AI follows my prompt and probably multiple iterations of it. Whereas if you wanted frame at a time, then it would be considerably more expensive. I think Damien actually ran a cost a cost thing because he gets sometimes drive drive by comments on his Star Wars iClone production of oh this would look better with AI, which of course Damien does not agree with that at all. But he actually just as a kind of a to humor himself ran a calculation of if I had done this with AI, with the number of frames of film that are involved with this, and I mean the cost was just astronomical, you know. Yeah. Tracy Harwood 34:28 So, but if this is done with style transfer, that's a slightly different thing, isn't it? Phil Rice 34:36 Well, this isn't. Tracy Harwood 34:38 This is what Phil Rice 34:38 I'm saying. Tracy Harwood 34:39 I think this is. No, I don't. Phil Rice 34:40 No, this was AI generated video. He had guidance images, at least at the beginning and the end. And I have not fired up AI Studio yet from Realusion. I've got it installed, but I haven't fired it up yet. So, but I believe that the main thing is beginning, end, and the prompt guides what happens in between those. And then AI studio helps keep the character consistent, helps keep certain environment things consistent, and it does a very good job at that, from what I can tell. Tracy Harwood 35:09 Right. Okay. Phil Rice 35:11 But style transfer to me would be more of that. It's it's actually interpreting every frame of what is submitted to it, like style transfer on an image, you give it an image, and then it upgrades the look of the image, right? So to do that for film, if it's 30 frames a second, you got to do that for 30 frames for every second of film. And I don't think that's what happened here because I don't see any indication of a tool involved where he would be scripting or keyframing any animation at all. So I have to think that is prompt driven. Tracy Harwood 35:42 Yes, I agree with you. I think I think that is prompt driven. But I but I did read where that style transfer is can be part of the process. Phil Rice 35:49 It can, but it is extraordinarily expensive, especially given that the the the outcome is less than certain. It's also Pretty time-consuming. AI-generated video has gotten to where it can be pretty quick. Tracy Harwood 36:06 Well, do you know? Especially if Phil Rice 36:07 you're not trying to do it locally, if you're relying on a cloud cloud muscle for it, AI-generated video clips, like through what was the one that I used for that that film last year? Runway. Yeah, it's pretty fast. You know, all things considered, it's pretty quick. I've not messed with Seed Dance yet. I know that Tracy Harwood 36:29 I've not messed with it either. Phil Rice 36:30 There's there's some people that I I interact with that are in Machinima on Facebook that have played with Seed Dance. His name is slipping my mind now, but McCallum, whatever his his Speaker 2 36:46 name is, yeah, yeah, Phil Rice 36:48 yeah, yeah, Tracy Harwood 36:49 yeah. Phil Rice 36:49 He's he's played with C Dents quite a bit and gets gets really good results out of it. As far as consistency goes, one of the things that's a challenge in even in guided AI video is to get that, I mean, precision level of consistency, especially on the character. That's been kind of a holy grail for a while, and when you just leave it to the AI model, it can go astray really fast. Like even in the course of one clip, it won't hold, and I think one one area where John has demonstrated real success here with AI Studio is with that consistency Tracy Harwood 37:30 absolutely Phil Rice 37:31 metric. It's it's it's flawless as far as I can tell, and that is possible with other tools, but it is very difficult. It is very challenging and requires a lot of kind of tinkering and configuring the the the generation model exactly right. And I think the one of the problems that they're out to solve with this AI studio is to make that an accessible process with a user interface and just nice and a stable thing that you can use, I think that's part of what this is demoing. Tracy Harwood 38:04 Yes, and Phil Rice 38:05 in that regard, it's it's just it's best in class. I mean, it really is. Again, I had assumed that there was a lot of control over the animations, but given that there wasn't, it's like like I said, it's mixed feelings. Part of me is like, ah, you know, I would have liked to seen that, but also, well, I mean, there wasn't anything weird there, right? There were no six-fingered people or, you know, strange quirks. Now, how many of the passes had to be discarded for those? We won't know, right? But, but I have a feeling it's probably pretty low for a couple of reasons. One, AI Studio helps stabilize that, and two, I think it's safe to say John's probably a pretty decent prompter. And I have long said that the the kingdom of AI really does belong to the people who master the language of prompting, that that is key. It won't make something like this for someone who doesn't know what they're doing with prompting. It will not. And all the good AI stuff that we see and that we'll probably be reviewing together here looks like it might take us only about a month to get enough films to do an episode on that. Tracy Harwood 39:22 Absolutely, all Phil Rice 39:23 of them do this in some way. Tracy Harwood 39:25 Yes, Phil Rice 39:26 they do what AI Studio is helping John do. It's just much harder for them. I think that this makes it, it opens it up to a new audience of people who aren't necessarily tinkerers and messing around with Python scripts and any of that stuff, you can just launch this program and supply the media, and it will it will do its thing. Now, how perfect it is on each pass, I don't know yet. Maybe I'll have some feedback on that when I if and when I play with it. But Tracy Harwood 39:57 yeah, yeah, Phil Rice 39:58 I think it does make it easier. I think there's no question about that. So, I noticed the lack of footprints too. I know I've just complimented John's prompting, which is excellent, I'm sure. But that could be fixed with a simple prompt. It may just be something that he didn't notice. There's one scene right near the end where he's walking away from the the water with the creature, and they're walking through sand and there's no footprints being left as they walk. Little things like that. When you're in the weeds on a movie, you sometimes don't notice those things. I remember. Tracy Harwood 40:31 Well, they're minor, aren't they? Really, and they are Phil Rice 40:34 minor. Tracy Harwood 40:34 Yeah, Phil Rice 40:35 but but many. I remember many years ago, I put out a film. I don't remember which film it was, but I was really, I was really pleased with it. And then Tom Jantal contacted me privately and says, "Where's the shadows? I was like, "What? Great looking film, but there's no shadows. And sure enough, he was right. I just completely neglected it because I just hadn't looked for it, so yeah, those things happen. I don't think that's a flaw in the in the workflow of the software. It's just one of those things that you got to, you know, when you're when you're putting together an animation, there's a lot of stuff to watch, you know. So Tracy Harwood 41:16 absolutely. Phil Rice 41:17 As far as the the let's say the filmmaking itself-the the one part that grated on me, the only part really-is the overuse of crossfades. This is particularly true in the opening half of the movie. That just didn't really make sense to me Narratively, just way too much crossfades between scenes when a cut or something would have done, and I've been asking myself like, why would that? Why would it do that? Because I don't. The thing is, is I don't yet know if that was something that Seed Dance served up to him, or if that was done in post production in the video editing. I think the latter. I suspect it was done in post that there were clips. So it's a matter of okay. A lot of times, you're when you're doing AI generated video, you're dealing with small clips, you know, 1015, maybe 20 seconds, maybe at the most, and so to have that many cuts sometimes when you're not changing who's talking or when you're not when there's no reason to change the direction of the camera, you got to figure out how to get from this shot where the head is right here to this shot where the head is right here. That if you do that without any transition, it's a jump cut and it feels weird. So it was probably just a matter of that's the lesser of two evils. If okay, well we could just crossfade those. Tracy Harwood 42:57 Right. Phil Rice 42:57 I don't know for sure, but it was just too much, too much of that to where that was that was a little jarring to me, and that's when I started to suspect. Again, I I watched it before I read the the how to description. That's when I started to suspect. Okay, this is not iClone to AI because in iClone you can have much longer scenes than that, right? If I had a three-minute scene, I could render the whole three-minute scene and have it be out. So this feels more like I had a 15-second clip and then another 15-second clip, and I had to join them together somehow. And so a quick quick crossfade probably just ended up looking better than a hard cut for those, but it's not perfect. You don't like it, but it's better than the alternative. That that may be what it comes down to. I don't know. Maybe John will comment and let us know on that. But I'm not tearing it apart for that. It's just that that jumped out to me a little bit in terms of just purely from a filmmaking point of view. You know, crossfades kind of have a a pretty specific place in the cinematic language, and this seemed to to kind of violate that a bit. So, the other thing that that jumped out to me a little bit was the AI voice. the the The voice of the human was clearly AI. Tracy Harwood 44:23 Oh right. Okay. Clearly, Phil Rice 44:25 yeah. And okay. So if that's a tough one, because if if the point of the project was make as much of this AI as possible, then what else are you going to do? You know. But That portion of the audio, the voice, gave this gave me an AI slop shiver more than anything did visually. Visually, I didn't get that off it. Again, I could tell that it was AI, but I thought. This is good, you know. It's it's well done, and again, we're going to be covering some more examples that we've started to find. That it's like, okay, this is pretty impressive. This falls in that category for sure. It does. Yeah, the AI voice. Yeah, you can't. I don't know. Maybe some people are just really easily fooled by that and don't notice it. But I can be in one room and my wife in the other, and I hear pull up Instagram, and I hear a reel going, and I'm just like, ah, AI voice, and it's it's just so pervasive now, you know? Tracy Harwood 45:31 Yeah, it is. Phil Rice 45:32 So anyway, that that would have again if it violated the the rules that he set up for himself on I want to make this with all the AI tools as as a proof of concept, well, then you got to do what you got to do. But if if it wasn't that, if it was I'm going to make a film, but use these tools, then you know, just use a voice actor for that or something because it just it really for me it just really took me out of it. I I have a question too and and don't know the answer about how the sound and music was handled. I suspect that the sound was rendered with the clip, the sound effects that includes ambient and sound effects sounds. I think that's almost certainly the case. Tracy Harwood 46:18 You think, yeah, Phil Rice 46:20 yeah. But the music I'm less certain about. He doesn't reference one of the standalone music AI tools like Suno. So, but it sounds like Suno type of music. Tracy Harwood 46:32 Okay, Phil Rice 46:33 and if if the music were confined to the clip, then I would know it came from C Dance. But there's points where the music goes over from one clip to the other. Tracy Harwood 46:43 Yeah, that Phil Rice 46:44 could just be some clever editing. But then you would have had to separate the sound effects track from the music track, and I don't think any AI rendered video gives you multi-track audio yet. So I don't know how that was done exactly, but it did. It did sound like AI music, but it was it was decent. You know, it was. Oh, it Tracy Harwood 47:06 worked. It definitely felt Phil Rice 47:08 like it worked. It followed the emotional flow of the story precisely well. Tracy Harwood 47:14 Yeah. Phil Rice 47:14 When the action picked up, when he was sad, when you know, it it really it was appropriate score. How it was created, I'm not exactly sure. You know, if if if we didn't have any information about the tools used, I would absolutely just assume it was Suno. But because he doesn't mention it, and he's careful to mention all the other things, I'm not sure. And again, I haven't played with Cdnce at all, so I know that it can do score, that it's it's something you prompt it for, and it will will do that, and that it does a competent job. But I don't know if that's the case. That I don't know how he spanned it across multiple clips. So again, John, if you're listening, solve that mystery for us. So, um, yeah, I mean, this is one of the better AI involved videos that I think I've seen, and the consistency and the control is is really the the star of this, I think. And yeah, I definitely think that this was made with dual purposes in mind. I think it's a it's a proof of concept to help a new product that they're very excited about. But John has done these over the years, and he always puts more thought to it than just "Hey, let's make a marketing video. You know, I can't name them off the top of my head, but I can think of two other ones that he has produced that really had that same dual dual functionality going on. It was to show off the latest. One of them was to show off the latest facial animations of character creator three or four-it's several years ago-and it was just a guy sitting in a chair reciting a poem. I can't remember the poet's name. It's a 20th century poet that John is a fan of, and Ricky's a fan of too. And I'm-I'm totally blanking on the name, but it was just a it was a poetry reading, but but and I think that was picked because it he personally likes you know that poem and the and the voice performance was very well done. I'm not sure who did it. It was not AI, but but also it was released as hey check out what you can do with facial expression now you know and there's another one that he did that was like something to do with a radio operator on the roof of a townhouse or something like that historical setting black and white so he's done these before and they are full purpose but he also. He puts a little of himself into them and makes them personal art projects as well, and I think that this succeeded in that regard. Tracy Harwood 50:09 Fair enough. Fair enough. Phil Rice 50:10 Yeah, I think that's everything I had to say about it. Tracy Harwood 50:14 Yeah, yeah, great, great discussion. I mean, did it? Do you think because the one that we looked at as well, where where the creator was very clearly doing a tech demo as well, was the gluons. To me, this doesn't quite hold to that level of the gluons. Don't know what you think. Phil Rice 50:41 Well, the motivation is a little different, obviously, because because Tracy Harwood 50:44 yeah, for sure, the Phil Rice 50:45 Glorons guy doesn't doesn't have any personal stake in the success of those ventures. For him, it was a proof of concept for himself, right? Tracy Harwood 50:55 Absolutely, Phil Rice 50:56 John's doing it for the world, so it's a little different. Yeah, that's a tough call. Gloron's is really good, and his his his following stuff too. There's one sci-fi thing that he does, and I mean, he is leader of the pack on that, as far as I'm concerned. And he's doing all this character consistency stuff, Tracy Harwood 51:14 yeah, Phil Rice 51:15 without AI studio. So Tracy Harwood 51:18 indeed, Phil Rice 51:20 I hope that he finds out about that tool because I think he'd find it. You know, maybe it'd be a way Tracy Harwood 51:26 to to Phil Rice 51:27 make his life a little easier. Tracy Harwood 51:31 If indeed that's his goal. Phil Rice 51:33 Yeah, yeah, that's true. That's true. Tracy Harwood 51:37 Marvelous. All Phil Rice 51:38 right. Well, audience, what did you guys think? Do you have anything else to add, Tracy? Tracy Harwood 51:42 No, no. I think it was a great discussion. Thank you very much, Phil. Phil Rice 51:45 Yeah, super. We'd love to hear what you think in the audience. Our email is talk@completelymichinma.com, or you could drop us a comment wherever you watch this video or listen to the podcast. We do read all the emails and comments, and would love to hear from you. So my name is Phil Rice, and on behalf of my co-host Tracy Harwood, we will see you next time. Tracy Harwood 52:05 Goodbye.