Google Just Dropped Nano Banana for Video

Añadido: 2026-05-23

[00:00:00]Google's next big AI video model is here and I have some thoughts. We're doubling down on our AI agent bet and the future of AI video production isn't seed dance or even Google VO, but a brand new tool that is ushering in a new era in creativity. We have all of that and more in this week's episode of AI Film News.

[00:00:23]Thank you so much for joining. My name's Caleb. Now, before I get going, I want to give a hat tip to the Can International Film Festival for featuring Dave Clark, who is the co-founder of Promise and also the instructor of our AI advertising course here at Curious Refuge. He recently spoke at the festival and is also one of the finalists for the runway AI film festival, which is happening in just a few weeks. Now, the news of course that we're going to focus on is Google Omni Flash. What is Google Omni Flash? Well, at this time, the best way to think about it is kind of like nano banana for video. You have the ability to input different media types, whether it's images, videos, text strings, or audio clips, and output a video. That is right now. In the future, they say that you will be able to input anything and then output anything and return. But right now, it's only for video. Now, there are some limitations that you should know about. The first is you only have the ability to output a video in 10 seconds at 720p. It's also only available for people who are subscribed to a Google AI Plus Pro or Ultra plan. Now, as the name implies, this is a flash version, meaning a very quick version of the tool. I talked with the DeepMind team earlier this week before this tool actually released and they were mentioning how it does have the ability to go beyond the current limitations, but obviously for security and data reasons, they're releasing the faster version at this time. But with all of that being said, we were very curious to test out the tool for ourselves. So, let's hop into the tool and see how it stacks up. To use Google Omni, you do have the ability to generate it inside of Google Gemini directly, but you also can use Google Flow, which I'm going to use because I like the platform. So, all you have to do is go to the new project button here, and at the bottom, you'll see we have a setting feature. If you go ahead and select that, you can actually click this menu here and make sure you are selecting Omni Flash as opposed to Google VM. So, we'll select Omni Flash here. we want to generate a video and then we'll select ingredients because we want to bring in multiple assets together. Now from here we can bring in our assets. All we have to do is select the plus icon and go to upload media and select the asset that you want to bring in. For our example, I'm going to use this drone footage here and basically I want to prompt for a sky beam to hit the towers in the middle. So we'll go ahead and bring that into Google Flow. And I'm not going to read out the entire prompt, but basically be prompted in for the drone scene to have a sky beam that comes down and destroys the buildings.

[00:03:00]And we'll go ahead and click render. So after a few minutes, we got this video clip from Gemini Omni.

[00:03:14]Okay. Uh so not the most photorealistic result. It almost looks like there's clip art on top of a video. Now, let's of course compare that against Seedance 2.0 Omni, which is kind of the industry standard Omni tool at this point in time.

[00:03:45]Okay, it's better, but not good. So, there are definitely limitations to these tools. Now, let's speedrun through a few other examples here. So, I inputed this reference footage of this man walking in the desert, and I wanted to prompt for an explosion in the background, and that gave us this inside of Google Omni Flash.

[00:04:10]Okay, so not very good. We got this result from Cedance 2.0.

[00:04:24]And this one actually looks much better.

[00:04:26]And uh that guy is awesome because he walks towards explosions. He's even cooler than Tom Cruz running away from explosions. So we wanted to do another example here where we have multiple reference pieces. So I inputed two different reference images. reference number one of this man and then reference number two where we swapped his head for an alien. And then we inputed a third reference video showcasing a transformation effect. And essentially we prompted for the man to turn into an alien in a similar way to the transformation effect that we uploaded. Now inside of Gemini Omni Flash we got this.

[00:05:09]And that's not an editing mistake. It literally cut out right before it did the transformation. And I guess it went with a like a paper aesthetic like maybe he's a piñata man. Not entirely sure what's going on there, but that is not what we're looking for. And I should note that there's some really noticeable diagonal banding happening. It's one of the larger problems inside of the Google video generation tools and it seems to be present inside of Omni. Now we did the same generation inside of Seed Dance and it gave us this video clip.

[00:05:47]Okay, not too bad. I think it did a pretty good job. I wouldn't say it's deeply realistic, but the fact that it did all of that from a single prompt is pretty impressive. Now, we were also curious to test out the trick where you upload a picture of a person and then black video with audio underneath to create lip syncing. And so we have this image of I guess that's kind of a picture of me sitting on a park bench here. And we wanted to include this audio clip.

[00:06:13]The rain always has a way of finding me better than anyone else apparently.

[00:06:18]Okay, that's a very bad AI clone of my voice, but you get the idea. So inside of Gemini Omni Flash, we got this.

[00:06:27]The rain always has a way of finding me.

[00:06:30]Better than anyone else, apparently.

[00:06:33]Okay, so there's some things to like about this. Obviously, it's more or less the same scene, but there's a lot of just kind of strange hypersharpening happening. The background has a texture on it that looks very unrealistic. And of course, that's not the same voice that we uploaded. It changed the voice to something else. So, not really the best result. Now, let's compare that to the same generation inside of Cedance 2.0.

[00:07:01]Uh, the rain always has a way of finding me. Better than anyone else, apparently.

[00:07:08]Okay, again, it's not a great generation, but it is far more impressive than the version from Gemini Omni. So, let's do a few other comparisons here. So, I have some reference footage of this woman walking through a salt flat, and we prompted for a spaceship to crash in the background inside of Gemini Omni Flash. It generated this result. Let me know if you see anything familiar here.

[00:07:45]Uh so yeah, Star Destroyer hitting the ground there. And this is the same result that we got from Seed Dance.

[00:08:03]Okay, again, not perfect and there's kind of some weird stuttering happening which you would have to fix inside of a tool like Topaz Video AI. Uh, but from a strict IP perspective, I definitely think you would want to go with the seed dance route here, which I did not think I would be saying inside of this video because I would have expected that Google Omni Flash would have had more IP restrictions, which is kind of interesting. Now, let's go through three more examples here. So, I have this reference footage of this truck driving away from these asteroids. This was generated inside of Seed Dance, but I wanted to change the asteroids to meatballs.

[00:08:39]And Gemini gave us this, which is not the same video. There's some cool stuff happening. I like the physics on the spaghetti.

[00:08:47]It's really funny. Uh, but it's not exactly the same video. Now, let's compare that against Sea Dance, which gave us this result here.

[00:08:56]There's definitely some physics that are off with the meatballs, but I think overall it's still doing a better job.

[00:09:02]So, again, props to Seed Dance again.

[00:09:04]And I wanted to push really deeply into a cinematic visual effects shot here.

[00:09:09]And so I have some reference footage of basically just kind of handheld out the car window here. And I wanted to prompt for a creature to come through the sky.

[00:09:18]It was pretty nondescript. It could just be any creature. And we prompted inside of Google Omni. And that gave us this, which it's a sky whale, which is uh kind of cool. I wasn't specific about the feature, so why not a sky whale? And inside of Seed Dance, it generated this video clip here, which uh is frankly much more impressive. Like the way in which the clouds just kind of billow around the dragon, it's really really impressive.

[00:09:50]So I think Sea Dance again did a better job. And this final example is actually pretty interesting and it will help me illustrate why I actually think this is a part of something much larger that is on the horizon. So I have this reference video here from art list of this man walking down the street and I prompted inside of the Omni models that I wanted to change the environment to the 1920s.

[00:10:14]So inside of Cling it gave us this result. So, there's some aspects that look like the 1920s, like that car, I guess, is kind of 1920sesque, but I'm pretty sure they did not have line scooters back then. So, uh I don't know if it did the best job. Now, we did the same thing inside of Luma, and that gave us this shot here, which I do like that it made the man wear clothes from the 1920s, but again, we have that Lime scooter, so that's not exactly what we're looking for. And I'm fairly certain they did not have Mini Coopers back in the 1920s, but I do need to do my research to really figure out if that is true or not. This is the same generation inside of Seedance 2.0.

[00:10:55]So, you know, the overall scene, I would say, generally looks like it's from the 1920s, but again, some of the cars in the background are much more modern. And of course, it changed the scene entirely, so it didn't do the best job.

[00:11:08]And this is the result from Google Omni Flash. And this embodies more of that 1920s aesthetic. The details look much more like they're from the 1920s. So intellectually it did a better job, but the overall physics and dynamics are not quite what you would want from an actual professional project. So with all of that being said, I want to talk about our quick take on Google Omni Flash. The short takeaway is it's really not a tool for professional use. And the DeepMind team specifically said that it was much more for general content creators and just kind of everyday creativity. This is an early iteration of this tool and it's really not in the same league as other omni models like Seedance 2.0. So the thing that makes this tool interesting at least in the future if not right now is the fact that it's built on Google's data infrastructure.

[00:11:59]What I mean by that is hypothetically this model should more intelligently understand the world and what you're trying to generate rather than understanding pattern recognition. You should in the future be able to use this tool to do everything from intelligent rotoscoping to changing the era in which your film is shot. And because it's pulling from real-time data and the larger world model that is encompassing the entire Google AI ecosystem, I wouldn't be sleeping on Google. Even though it does seem that they're slightly behind seed dance at this time, I think that this model is going to be incredibly powerful in future iterations and it is going to outperform other tools that only go through model training and pattern recognition alone.

[00:12:45]And I should note that a lot of people said that this was supposed to be Google V4 and it is not Google V4. Google V4 is either coming soon or not coming at all.

[00:12:54]But Google Flash Omni is the latest version of AI video from Google. So you can play around with it yourself if you have at least a pro subscription to the platform. Now that is not the only feature that Google announced and unveiled at Google IO. They also came out with a new feature inside of Google Genie 3 that allows you to essentially create an interactive world from Google Street View. If you're not familiar already, Google Genie3 basically gives you the ability to upload an image or type in a prompt and create an interactive 3D world, almost like a video game. To use this latest version of Genie 3, you do have to have a Google Ultra subscription. So, to use it, all you have to do is go to the Genie3 website and let's go ahead and click on that try in Genie button here and we'll go ahead and click explore now. So, this is not new. This has been around for a while. It's just one of the additional location features that makes it kind of interesting. So, let's go ahead and select create your own here. And at the very bottom, we're going to select this brand new Google Maps button. So, for this example, let's do the Space Needle in Seattle. Go ahead and select and we'll select use this location. And after a few minutes, it generated this 3D world that we have the ability to navigate. And uh yeah, we're just a cat walking around this Space Needle here.

[00:14:18]And I think that cat's like way too big.

[00:14:20]It's like lion size. Now, tell me if you see something missing from this video here. Uh I'll give you a hint. It's what we actually prompted for. The actual Space Needle is missing from this 3D world, which is which is really funny. Uh, so it is interesting that it's generating a world in 3D that you can interact with it. But of course, there's still some gaps. It's the early days. It's not to say that world models will not be interesting in the future. They 100% are a part of the future of creativity. It's just that Genie 3 at this point uh maybe doesn't fully understand the context of where the generation is happening. I'm also super excited to announce that our long awaited advanced AI filmmaking 2.0 0 course is finally here. The course goes on sale on May 27th. We've partnered up with Cavin Cardardoza, also known as Caven the Kid. He's one of the world's best AI filmmakers, and this is the latest iteration of his AI filmmaking course. Inside the training, he will get into workflows and systems that you can use inside your creative projects that he has not shown up to this point. This is the best way to take a look at professional workflows used by one of the very best AI artists in the world. I'm super excited for it. You can learn more by clicking the link below this video and join the wait list today.

[00:15:47]Next up, we came across an AI image tool that says it's an open-source competitor to Google Nano Banana. So, naturally, we were very curious. The brand new tool is called Hydream and there's a hugging face example online where you can play around with it for yourself or you can just download it and install it to your computer. So, I'm running on hugging face here and I apologize for the UI in advance. I have no idea why their UI looks terrible, but it just does. So, we'll just go with it. And I won't bore you with the prompt. Basically, we tried to create a prompt of a diagram of how you create sushi. And you can adjust all the settings that you would want to adjust inside an AI image tool. And we'll go ahead and click generate. And after a few seconds, we got this image here, which uh looks okay at a quick glance. And then you start looking at the text.

[00:16:39]Ever. I have no idea what's happening.

[00:16:41]And that sushi looks disgusting. I have no idea what that is, but it kind of looks like baloney. And the actual sushi roll, I think, has macaroni and cheese in it, which maybe that's good, but I I doubt it. And we prompted for the same thing inside of Nano Banana. And that gave us this image, which looks really good. It's really, really impressive, frankly. We also prompted for this man on a dirt road. And it looks good, but if you think about it, the road is facing the wrong direction. Like, he should just have a field behind him, unless he was at that moment right at an intersection, which is probably not accurate. And inside of Nano Banana, we got this image. So, again, Nano Banana is much better. We prompted for a fantasy scene with this elf here and this like evil demon character. They gave us this from Hydream. It's It's okay. Uh whereas we got this from Nano Banana 2, which looks like basically photorealistic concept art. We prompted for a 1985 video game on a Mac. It generated this. Again, the text looks just like super weird. Whereas this is what we got from Nano Banana. And I love the Goonies poster in the background.

[00:17:49]That is just like 10 out of 10. Great job, Nano. So, long story short, is High Dream a real competitor to Nano Banana?

[00:17:57]Absolutely not. I don't think it's very good at all. So, moving on to our next conversation point. Let's have a conversation about AI agents. It seems like every single tool company in the world is coming out with their own AI agents, and they can do a lot of things that are really helpful, especially when it comes to workflows and productivity in a more traditional professional sense. But there have been a lot of creative agents that have been popping up recently. We've been very skeptical because we've tested all of these agents and long story short, they haven't been very helpful. They've been slow and ultimately not that creative or interesting. And so we put an open call out to the community last week where I said all you had to do is share one example of utilizing an AI creative agent that is better than you simply going to the tool yourself. and I offered $100 to the first person to give us a really good example that we agree with. Now, we took all of the submissions last week and tested every single recommendation in the comments.

[00:19:01]And here's a quick overview of the results. First off, we had quite a few people say that AI agents are really good for double-checking character continuity inside of your projects. The idea is you could create a board that has your character or your environment and then use an agent to cross reference images that it creates to give you better continuity. So I decided to test that out for myself. So for this example I'm utilizing Luma agents, but of course you can use any agentic creative tool that you want. I'm not specifically trying to single out Luma by any stretch here. Okay, so we have two images here.

[00:19:38]We have a picture of me and we have a coffee shop. And basically I went to the agent and I said create a wide shot of the man inside of the coffee shop drinking coffee at the counter and to work as quickly as possible and create the image in 16x9 and then to doublech checkck that the character is the character that I uploaded. So I'm trying to use character continuity here. And so we sent the request and the entire process took exactly 2 minutes and 41 seconds and that gave us this image here which is kind of funny. It said that it's 94.9% a biometric match and that the threshold was 85%. And if I zoom in here, I got to say like it's close. Like maybe 95, but I would say probably a little less. Like I would say that this character is perhaps a distant cousin or something like that. And uh there's just like a lot of texturing on the skin that makes this guy look much older than me, which is kind of funny. And I should note that the shirt that he's wearing is absolutely not the same shirt. So that's not the same character outfit that we're looking for. And again, it took 2 minutes and 41 seconds to pull that off.

[00:20:48]So I did the exact same thing inside of Magnific. I uploaded both of those images and asked for the exact same shot. Only this time, I'm going to be the one curating. And of course, you have the ability to do four outputs at a time. You can change the resolution.

[00:21:02]I'll keep it at 1K for this example. And they also do have an unlimited model, but I turned that off for the example.

[00:21:08]And we can just go ahead and click generate. We can click generate a few times actually. So we can hit it once, load, and go ahead and click generate again. And you can go ahead and hit generate again. And so essentially, as fast as we can click that generate button, we have images that are being created. For example, we have this image here, which is me wearing the exact same outfit. And that just looks much more like me. It's just a much more realistic image. That definitely is the coffee shop here. So, it's doing a much much better job. Now, some of these results are not very good. For example, this isn't exactly right, and it just added in a table here, which is not what we're looking for. And some of the generations just look silly, like the lighting here.

[00:21:51]The proportions are not what we're looking for. But some of them are actually decent. And again, it's AI.

[00:21:57]It's all about iteration, not one thing at a time. And the problem is most agentic workflows use a waterfall project management methodology which means one task happens then the next then the next rather than iterating a bunch in each task. The problem is it really just slows down the production process and ultimately in my opinion does not do a better job. So everybody that said it's better for continuity unless we're missing something here I actually don't think it does a better job. We also had other people say that utilizing Runway's new agent feature allows you to upload your script and get a film back in return that is just as good as the one that you would create on your own and it's so much faster. And so we wanted to test it out for ourselves.

[00:22:41]And so we went into Runway Agents and we pasted in a script for a film that we've actually demoed in other Agentic Workflow videos in the past. Basically, it's about a man who's sitting at home and he's watching TV and something happens and he has to go back to the laboratory and then he gets sent back in time. Okay, so very simple story and we typed it in and inside of Runway it generated the characters, it generated the sets and ultimately we can click on that generate video button and it generated a video automatically. And so let's see if it did a good job.

[00:23:13]>> The story you are about to see is true.

[00:23:15]The names have been changed to protect the innocent. was a case that defied all logic. A case that would change the course of history. A case that would reveal the existence of a secret organization created by Christopher Nolan.

[00:23:32]So, what's so funny is we did not prompt for it to create a trailer. We wanted it to actually create the scene and it just made this very, very quick trailer.

[00:23:41]Obviously, there's very clear continuity issues. Some aspects of it are interesting, like when the guy hits the TV remote. Those two shots, I do think it works well together, but ultimately there's inconsistency in the style. It randomly goes to grayscale in the time machine scene. So, I don't really know what's happening there. Ultimately, it's just not an efficient workflow compared to if you were utilizing it yourself.

[00:24:04]So, again, runway agent probably not that helpful. And the final suggestion that we saw come through was to utilize Utopi AI, which is an online tool that's semi-agentic. Basically, it's very similar to these other tools and workflows. You have the ability to upload a script and it kind of walks you through a procedural workflow to ultimately get the creative project that you're looking for. For that example, we uploaded the same script and it gave us this.

[00:24:30]>> Imagine a world where our choices were no longer our own. Screaming for the one thing you can never replace, your time.

[00:24:37]Welcome to the noise. Dave called. I got to go in >> this late.

[00:24:46]>> I know, but whatever it is, it must be big. I've never heard him so panicked.

[00:24:50]>> Well, tell Dave your wife expects you to stay home Friday.

[00:24:53]>> You and me both. All right. Bye.

[00:24:56]>> Love you, >> Dave. I was halfway through dinner. You said the cooling manifold was redlinining.

[00:25:04]>> Then what's red line? Logic loop in the predictive engine. I lied. We've been trying to push the QC core super position state >> right to minimize deco.

[00:25:11]>> The textbook's wrong. We didn't minimize it. We fold it. I was running an acoustal line test. Monitor the transistors for background vibration.

[00:25:18]The monitor started spiking. Watch the rolling feed. I'm not touching any silent. That's got to be a sensor latency issue.

[00:25:25]>> I checked the atomic clock. There is no lag. Dave, what's happening? I don't know. The entanglement must be cascading.

[00:25:31]>> Dave, look at me.

[00:25:33]>> The phase space is shattering. We're seeing the probability ripples.

[00:25:38]>> Shut it down. Kill the core.

[00:25:41]>> I can't.

[00:25:47]Oh my god.

[00:25:51]Okay.

[00:25:54]So So it's not good. It's not It's not good, but it's so bad that it's good. I I want to watch so many more of these AI films. Uh this sounds like a you know a late Friday night watch session. I just want to watch uh so many of these absurd AI projects. Uh so again this is not a good agentic workflow. The project is not good at all. There's some things that look interesting. It kind of hints at something cool and ultimately agentic AI workflows do not seem to be impressive in their current form. They will likely be awesome in the future.

[00:26:29]Right now not very impressive. And so we're going to double down on the challenge from last week. We are now going to give $200 to anybody in the comments that can give us an actual agentic creative AI workflow that is actually better than if you just went in and prompted for yourself. Now, shifting gears, I want to talk about an update that came from the team at Quinn. Long story short, one of the biggest problems in the process of generating creative AI assets is simply the data restriction, right? There are only so many GPUs in the world. And so, if the current models remain the same, it's going to be very difficult to make large scale generations more affordable. We've seen time and time again that the process of putting together an AI film is getting more and more expensive. And so, the only workaround really is to make the algorithms faster and cheaper. The team at Quinn actually came out with a compression algorithm designed for images that can shrink an image down 32 times smaller than the original image.

[00:27:32]Long story short, if you can shrink down that image and then bake in AI data that can upscale it in the generation process, then you have the ability to generate a video that is 32 times smaller than the output format that you're looking for and still you can retain all of the detail and quality that you're looking for. The thing that makes this algorithm so interesting is it's actually good with the likeness of characters and text. And that's really important because you want to have all of those fine details and not get just the weird mush in details that you see in a lot of the AI video and image tools up to this point. So, we've talked a lot about the current version of AI video tools, whether it's image to video or omni tools. Now, I want to talk about the future of video production. The team at Odyssey released a brand new world model that generates AI video and audio in real time. So, this isn't hitting render and waiting. It's actually creating the action in real time along with all of the sound effects so that you can actually see the action as it happens and upload your own audio. You can edit what is happening and basically it loops and gives you the ability to edit your scene. This is the future of AI video production and it's really impressive. For example, we have this video clip of this woman on a subway.

[00:28:55]And again, this is generated in real time. They didn't render this and then upload it. It's actually happening in real time.

[00:29:02]>> I used to travel without looking back.

[00:29:05]Now I see the world in reverse.

[00:29:09]Memories haunt me.

[00:29:12]Regrets linger and haunt me.

[00:29:17]Okay. So, is that as good as Seed Dance or Google VO or Cling? Probably not. But the fact that it's generating in real time hints that in the future you will be able to edit that performance by just typing in a prompt or even audibly changing what you want that performance to look like. Think about it as essentially like a 3D film set that gives you the ability to direct the actual AI action that is happening. It's going to completely upend and change the way that we create AI films. Let's take a look at a few other examples here. So, we have this guy playing bongos. Again, these sound effects are generated in real time.

[00:30:01]Not too shabby. We have this video of a boat.

[00:30:15]And again, it looks pretty darn good.

[00:30:17]It's not perfect, but the fact that you can do real-time water simulations is nuts. Water in liquid is one of the hardest things to simulate because there's just so many particles that you have to run through your computational algorithms if you're going through a traditional VFX pipeline. And so the fact that you can use AI to pull this off so quickly is frankly crazy crazy impressive. Now you also have the ability to upload audio utilizing this tool. And here's a quick example >> to simulate it.

[00:30:48]Today we announce Star Child 1.

[00:30:55]So, it's not perfect, but the fact that you have the ability to upload the audio and prompt in real time for the action you want to see is really, really astounding. Again, I can't state how big of a deal this is. Imagine in the future you are working on a film project and inside of your editing timeline, you can simply loop and redirect the action that you want to see until it's exactly what you're looking for and then move on to the next clip. That's exactly the value proposition utilizing Star Child 1. And you can even see how this progresses into automatic films and creative entertainment experiences that are going to be unlike anything that we've consumed up to this point. Now, if that was the only announcement from Odyssey, it would be a pretty astounding week.

[00:31:39]But they also announced a real-time video game engine called Agora 1 that is somewhat of a competitor to Genie 3, but this one actually unlocks different gameplay mechanics. Long story short, it has the ability to generate interactive real-time simulation experiences with different gameplay mechanics. And the wild part is you can interact with other people inside of that world. You can actually play around with the game engine yourself. So, if you go over to the link below this video, you can click on the experience Agora 1 button.

[00:32:10]They've actually created a real-time game that's kind of similar to Golden Eye that you can play against other players online. I highly recommend checking it out for yourself. You'll find a link below this video. In Cedance news, we saw a rumor this last week that said Seedance 2.1 and Seedance 2.0 Mini are coming very, very soon, probably within the next few days. Seedance 2.1 is supposedly about 20% better than 2.0, which is already impressive because it's the best video model on the market. And Cance 2.0 Mini is going to be a lighter version that is going to cost about 7 cents per second. So, a much more affordable option versus the current version of Seed Dance. Now, I also heard a rumor that Seance 2.1 might be in 4K as well. That will be really helpful to have native generations directly in 4K because the only model that can do that up to this point is actually Cling. We also saw an update this last week where the team at Adobe has partnered up with Google to integrate Adobe's tools inside of Google Gemini. Long story short, you will have those tools directly inside of Gemini. So whether it's features like editing an image inside of Photoshop or editing it inside a video application that will be directly integrated inside of Gemini, it really is a monumental shift in the way in which we think about the process of putting our creative projects together. And we came across a really funny story this last week where an artist essentially uploaded a picture of a real Monae painting and said, "Hey, what do you guys think of this AI image that I created?" And there were so many funny responses. People said that it didn't look anywhere like a Monae and it's somebody trying to replicate his style or that it looks like high school level artwork. And you get the idea.

[00:34:01]People love to hate AI imagery and video, even if the quality is as good as other media sources that they've seen up to this point. It's really funny. I highly recommend checking out the text thread for yourself. Now, there are a lot of AI film events popping up around the industry. We have a free online workshop happening digitally on Tuesday the 26th at 11:00 a.m. Pacific time. We asked Auron Rabininoitz, who put together a really impressive AI superhero film, to come on to our speaker series and break down his process and talk about a few of the creative ways in which he was able to create that film. Now, on May 27th, the Curious Refuge community will be at AI on the Lot. Please join us at the 5:00 PM happy hour session where we would love to meet you and get to learn more about you as an artist. We're hosting a Curious Refuge meetup in Denver on June 6th and also a Curious Refuge meetup in San Francisco on May 28th. And that brings us to our AI films of the week.

[00:35:05]The first film that I want to shout out is called Stormchasers created by Dstudio. It's really impressive and it kind of has this kind of documentaryesque vibe and the actual tornado shot is really really impressive. Just a really good example of AI V effects in action.

[00:35:22]The next project that I want to shout out is called Zombie Scavenger from AI Movie Guy. It has really great visuals.

[00:35:30]There's a lot of action and complexity, and I think the creator did a really good job pulling it off. So, props to you. And then the final project I want to give a shout out to is called The Dream Station from Magu Directors and Gus V. Martin. I really love the visuals and the overall storytelling here, and it looks like it could be a fantastic ad campaign for an actual product. Thank you so much for watching this week's episode of AI Film News. As always, if you want to join the world's best AI filmmaking community, I highly recommend checking out membership for Curious Refuge by clicking the link below this video. And be sure to like and subscribe to get the latest AI workflows and tutorials directly here on the platform.

[00:36:13]Thank you so much for watching. I will see you in the next one.