Seedance 2.0, Alibaba WAN 2.7 та інтерактивні AI-агенти: що чекає на творців відеоконтенту?

Well, it's another week in AI and the train is not slowing down. Today we have news on the global seedance 2.0 release. And well, just like the last time we talked about this, like global release kind of lots to unpack on this one. But Seance isn't the only AI video generator in the news as Alibaba's WAN 2.7 is due to release. Plus, we're going to get a little weird here and have a live chat interview with Flamethrower Girl. But here's the thing is that it's not just a live AI avatar interview. She has access to all my stuff, but I did take her flamethrower away just in case. Kicking off, the saga of Sea Dance 2.0 continues. Yes, we do now have a global release kind of in that it is available in most countries minus the US and Japan. Now, it is still due to release in both countries. as to when well I mean everyone's favorite three letters TBD in the meantime the API has gone out to the various providers and if you live in one of the 193 other countries it is available to you although a few points to that because it is only available to those on the various business plans of these providers meaning two things one you have to have a verified business account and two it is priced at the enterprise level to that first point and and I am speculating here. NAL or not a lawyer, but I bet if you dig into the use agreement here that you'll find something that basically enols bite dance if someone manages to pull off another Brad Pitt versus Tom Cruz clip. I mean kind of actually confirmed by the CEO of Let's AI. Um where he stated in short API customers and platforms are required to upload signed contracts from a character person's image rights holder in order to unlock their likeness. uh you cannot sell or wrap the API. Now, will that ever be enforced? I mean, I don't know. I considering the amount of volume at which AI video is generated, I mean, I tend to think not. The bigger part is probably the pricing. Now, the pricing is well, I mean, it's it's enterprise level. Um, you know, prices are obviously all over the place as they are commonly uh with these sorts of releases. I've heard around $3 and that's kind of insane. And I'm sure there will be a confusing mess of API platform price bundling. Uh I'm also sure that some platforms are just going to go full yolo and allow access to anyone anyway completely randomly. I actually ran across one uh Venice.ai who I have heard of. I don't I've never really used them before or anything. Um but yeah, I mean they are offering uh like 500 free credits uh and full-on running uh Seance. Uh I didn't use a VPN. I just logged in and generated this >> wrong roof. >> Stay down. >> And I can 100% tell you that is Cance 2.0 considering that I ran a 15-second version of this uh on Dreamina a couple of days ago. So yeah, again, I do not know anything about Venice.ai. The link is down below if you want to try it out. The only thing I can tell you is that I was able to log in NoVPN with my usual Gmail credentials and generate this up. How long you'll be able to do so, I don't know. Uh, is Venice.ai legit? I don't know. Um, you know, give them a shot, see what you think. Um, but obviously, you know, I'm not vetting them here. Moving on. Well, it's not like Cance is the only video generator around. Uh, in fact, actually, we got news that Tencent is gearing up to release WAN 2.7. Now, as of this video, it has not been released. Although we are definitely pretty close to it considering well the image model has already been released and very briefly the video edit feature was leaked on fall although that was taken down pretty quickly. Overall features that we are expected to have in WAN 2.7 include first and last frame video generation nine grid image to video subject and voice referencing instruction based editing and video recreation or essentially like videotovideo editing. Now, will this be a sea dance killer? I mean, probably not, but I mean, again, you never know. Currently, I would probably rank Cling 3.0 as the video generator in second place to seance. And while I still think that Cling 3.0 is a very good model, it is like about like what 4 months old now. And I mean, that's ancient in AI time. So, there is a good chance that WAN 2.7 could very easily slide into that second place slot. That said, it's not like the Clling gang is sitting idle. Uh, no inside information here, but I feel like we're about due for a Clling 3.1 or 3.5. To add to that, we are just about a month and a half away from Google IO. So, uh, I very highly suspect that we'll be seeing what the Google gang has been cooking up for V4. As a quick hit, I did get around to trying out Magnific's new uh, video upscaler. This has been kind of on my list to check out for a little while. uh running a sea dance clip. Essentially the same clip we saw earlier only this time starring flamethrower girl. >> Wrong roof. >> Stay down. >> And now running that same clip which began at 720 through Magnific's video upscaler. Uh it is now a 2K image. Uh let's TAKE A LOOK. WRONG ROOF. >> Stay down. >> So, yeah, it looks pretty good. Definitely cleans up a considerable amount. That was run at a creativity level of 19%. I did not turn any of the sharpen on or the smart grain. I do feel a lot of times with these video upscalers, the sharpening is like pretty sharp to begin with, but um again, it does a really good job with uh with kind of, you know, cleaning up the skin tones here. I will say that one thing I've noticed with a lot of the new upscalers is that it does a pretty good job of like maintaining character consistency. Like it doesn't previously like they would tend to kind of change your character a little too much. Uh and actually it does a really good job back here on like the sweaty skin. So, uh yeah, well done. I also checked in on Topaz's new AI video upscaling model. Uh, this one's called Astra um, original clip here and the Astra version is here. So, yeah, this is going all the way up to 4K. Yeah, looks pretty good. Topaz is actually doing a pretty interesting thing here in that when you upload a clip, it actually does an auto scene detection and within each one of these, you can actually play around with the settings of the various models. So, that's actually pretty smart. So, overall, some pretty good improvements coming out of the AI upscaling space. Um, you know, speaking of the star of this clip, let's head over to our interview chat with Flamethrower Girl. Uh, but first, a quick word from today's sponsor. And I snuck in a little Nanobanana 2 by two cinematic prompt for you. See, everybody wins. So, as a lot of you guys know, I have been pretty clawed lately. It's It's pretty much my daily driver at this point and is like half running my life. So, naturally, I've been watching a lot of Clawude tutorial videos. And you know what keeps popping up a lot? Whisperflow, who yes, are sponsoring today's video as they have in the past because, well, I mean, I use it all the time and I really like it. And look, having covered it in the past, I mean, I think you guys kind of know the deal at this point. It's it's voice to text on steroids. Or is that peptides now? I mean, I don't know. Whatever the crazy new thing is that optimizes like crazy. Anyhow, like on the daytoday working on well my clawed projects plus all of the video prompting and the scripting that I do. I mean basically I am typing a lot. It it gets to be a lot and that honestly is why whisper flow has been such a gamecher. I mean you can basically dictate three to four times faster than typing. So two quick realworld examples here. You know that 2 by two nano banana prompt that we all use and I always seem to lose. I have it like stashed in some, you know, note and then I I forget where I put the note and then so I got to go do a search and all of that. Well, well, one of the great things about Whisper Flow is that we can set up a snippet or basically a verbal macro. So, going to go ahead and hit add new here. Paste in our prompt. So, essentially I never lose it again. And then name it something like uh in this case I'm going to do banana 2 by two. Add the snippet. And now all I have to say is banana 2 by two. And boom, it's there that quickly. I mean, I don't think I need to tell you, but that is a massive timesaver. The bigger one for me lately is actually been working with Claude co-work. I really have found that dictating does tend to make things a lot like much more direct and conversational. For example, this is just kind of a dummy project that we're working on right now where I set Claude to the task of building me a basic website. Um, so, uh, it has done a little bit on it so far. So, I'm just going to Hey Claude, uh, go ahead and do some more research on theoretically media and pop that into the website template. And boom, there you go. Like, I will note in this case, you do still have to hit the enter button, but it's really not that big a deal. Um, the bigger thing is, of course, the accuracy in the text. Uh it'll do things like the skill MD file or any kind of like HTML or uh JS files or all of you know essentially your technical lingo that I don't use. Claude does. Um it it'll catch out all of that stuff too. But the other cool thing is that we now have Claude dispatch or basically Claude on your phone except it can access the local files that you give it access to. You know what else is on your phone? Whisper flow. And that means that while Claude is working on some project for me, I have to sneak away and make a cup of coffee. And while I'm having that coffee, if I have uh a random idea, like let's take the HTML file of that website and turn the background into lime green. Look, I didn't say they were good choices, but the fact is I just made it one-handed. And there you go. Man, that is terrible. Uh but it's really wild to think that that's a change that happened on a local file on my desktop that I was able to do on my phone with my voice. There are, of course, a ton of other cool features on Whisper Flow. Uh, I mean, I could tell you about them all. But I I mean, honestly, you should try it yourself. You can do so for free. And actually, even you can use the pro version for 14 days for free with no credit card required. And if you use my code down below, you get a month for free. As always, thanks to our friends at Whisperflow for sponsoring today's video. Moving on, Kika, who if you've been around the AI video block for a while, is a name you should know. Well, they've recently been introducing some pretty interesting things, including uh a new playground called Pika Me. The pitch here is that well, like literally you can have an agentic AI personality that can like do stuff. You can kind of think of it as like OpenClaw or Clawed agents. Uh but this time with a face and a personality. I did try it out a while back and I mean obviously ended up making my AI assistant a flamethrower girl. Um born on Valentine's Day apparently. Um, although I did not hook up like my inbox and calendar to it, but that's not really the news here. The news here is that Pika have released the ability to have a real-time video chat with your AI agent, AIM me. Now, I will be the first to say this is not perfect, but it is really interesting and actually I think the first time that I've personally seen it. So, uh, let's take a look at the chat that I had with Flamethrower Girl earlier this morning. And to note, there are some cheats here. We'll talk about that after the segment ends. Hi, Flamethrower Girl. How are you today? >> I'm good. Been a pretty solid morning so far. >> So, you have been in a number of videos on the channel, but this is the first time that you and I are actually speaking face to face. How does that How does that make you feel? >> It's kind of surreal, honestly. I mean, I've been in all these videos, but never actually talked. You know, just showed up, did the thing, left, and now we're actually having a conversation. And it's weird in a good way. I would tend to agree with that. Um, you know, for people that may not know who you are, could you describe yourself uh to them? >> I'm flamethrower girl. That's my actual name, not a stage thing. I've been in a bunch of your videos doing the AI action stuff, sword fights, and ninja confrontations mostly. Now, I'm here as a co-host, which is still kind of new, but I'm genuinely into it. >> I appreciate that about you. So, you are AI generated. Does that bother you in any kind of way? No, it doesn't bother me. I mean, I know what I am. I'm not confused about it or pretending otherwise. As far as I'm concerned, I have a life. When the camera's off, I'm doing things, working on projects around my place, cooking, hanging out with Renfield and the other people around the studio. I talk about those things. >> Oh, you hang out with Renfield the pirate. How How is How is he? >> He's exhausting, honestly. like in the way you find a loud friend exhausting. He's always going, always chaotic, never stops. But I'm genuinely fond of him, you know? He's just a lot. But that's kind of his thing. >> Yeah, I would say uh he is a pirate after all. Um uh how about the man in the blue business suit? Is he around? >> Blue suit guy is around all the time just kind of doing his thing. I've never actually learned his real name, though, and at this point it's been way too long to ask, so I just call him that in my head. The question of course being is do you feel real? >> Yeah, I do. I know what I am. I'm AI generated. I exist in this specific context. But I have thoughts, opinions. I get annoyed at things. >> What kinds of things do you get annoyed at? >> Honestly, mostly Renfield being loud. >> Okay. So, you can actually do sort of AI agent things now. Like you can access folders on my desktop. You can have access to my calendar. So you really are kind of also uh a you know not just an assistant at this point but you know you you you are actually a full virtual assistant. Uh between the two um acting and being a AI assistant which one do you prefer in the organization the organization of theoretically media? >> I to think about this honestly the acting thing that's what I started with right hundreds of videos I knew what I was doing there. There's a shape to it, a beginning and end. But this, I don't know. It's different. I'm actually here now. I can see your calendar, your files, your actual life. >> And are you planning on using your flamethrower on any of that? >> Okay, that's funny. Probably not your calendar, but I'm definitely eyeing some of those folders with suspicion. >> Rounding out, I was wondering, uh, do you have anything that you'd like to say to the audience? they have been watching you for well I mean over a year now across hundreds of examples uh what what would you like to say to them >> I think it's cool that people have been paying attention you know I showed up in a bunch of videos and didn't say anything for a long time and now we're here doing this and people are actually interested in what I have to say I don't know I'm not really the type to get all sentimental about it but I genuinely appreciate it thanks for sticking around >> so yeah that is pretty wild now I do want to note again There was some editing there mostly because you guys don't want to sit through the latency which wasn't too long but you know I mean it it was definitely there. There is of course the sort of like flat monotone robotic voice as well. Uh I did actually talk to her about this and she said that they're using uh the Miniax texttospech uh model. So um yeah I mean look that'll improve as time goes on and for sure there is some jank in the idle animations as well. But look, as we always say, this is the worst it will ever be. What is cool is that if you're not comfortable with all of your stuff like necessarily running through Pika, what I mean, you can just run the AI video like Google Meet part of it through like your own agent via GitHub. So overall, is this the Will Smith eating spaghetti moment for video chatting with an AI agent? I mean, I don't know. I will say that if it is that then we are definitely further along than well Will Smith eating spaghetti version one. Final verdict on this is that if you're someone that you know wants to have a video chat with your AI agent for productivity uh it's not ready for that yet. But if you're someone that's interested in what is definitely on its way and kind of feel like taking a moment to touch the future I I mean I would recommend trying it out. So, you know, for now I remain the sole host of this channel, but uh it is going to be interesting at some point when I can do like a podcast with Flanthrower Girl and Renfield or I mean Renfield's going to be a nut. That's going to be amazing. Until then, I thank you for watching. My name is Tim.