Glif

With AI, we can do amazing things. For example, I can give myself super powers. Or I can make things appear that don't actually exist. Or I can turn myself into an attractive woman and you wouldn't even know it's me. We can create amazing custom motion graphics or AI influencers, but I don't want to learn all these different tools. That's like so mid. Exactly. I don't want to study 20 different models that change every week. I don't want to learn complex workflows with many different tools that each require its own subscription. I don't even want to edit videos or extract start and end frames. I think in the future AI agents will do this work for us. Agents that understand all these different models and tools and can pick and use them intelligently. Agents can also do the video editing like extracting frames, cutting videos, or stitching them together. Today, I'll show you a tool that does exactly that. You have agents for all kinds of use cases. agents to add special effects to videos, to create AI generated videos that are minutes instead of seconds long. Agents to create motion graphics and even animated presentations, and a lot more. The library is constantly growing with new capabilities. You only have to chat with the AI and the agent does all the work. It picks the correct models and tools. It generates video, audio, transitions, special effects. It stitches everything together and it has access to all the latest state-of-the-art models like Kling, Nano Banana Pro, ElevenLabs, Seedream, every model you can think of in a single tool without a subscription. In this video, I will show you some mind-blowing examples of what these agents can do. We will see how easy it is and if we run into any problems. We will take a look at the pricing and we will discuss how we can use this technology to actually make money and grow our brands. My name is Florian Walther and this is the AI tool corner where I review the latest AI software to find out which ones can actually improve our lives and businesses. The tool is called Glif. I will put a link into the video description and don't confuse this with GIF and please don't Google what a GIF is. When we scroll down here on the landing page, we can see that this indeed has access to all these different models. Open AI's models, Anthropic, Google's Gemini, um, China's DeepSeek, whatever, Perplexity, everything. All these different video, image, music, voice models are in here, and you get 10 credits every day for free. Now, heads up, this is not a lot. We will take a closer look at the pricing later, but at least you can try out some of the features for free. To create an account, click on the get started button up here. And then the dashboard looks something like this. We have access to agents and workflows. We will talk about the difference between these two later. First, we want to try out agents. And every agent here on this grid is an AI chat that has a specific purpose. Each of these has access to specific tools and models to do one specific task. And up here, we can filter agents by different categories or use the search field up here. They are also ordered by popularity. So, the most popular agents are at the top and the less popular ones are further down. We can even create our own agent, but we will take a look at that later. For now, let's see some examples of what we can do with that. The special effects you saw in the intro were created with the AI Hollywood agent. This one is really cool. Here we can see a previewer. If you focus very hard, you can be anything. Let me show you. It adds special effects and transitions between two clips from a prompt. We will try this out in a moment. By the way, I will put a link to all the agents I use in this video in the video description below if you want to try them out yourself. Let's try this out live together. So, I have prepared two video clips. They are very raw. I just recorded them with my camera and I didn't edit them. Let's take a look at them. Number one, ah, something is wrong with my stomach. Give me a second. And the other clip is me on the toilet. This is for a joke. Much better. And then I want to use a special effect to make my head explode. But since I didn't edit these video clips yet, there is some part in the beginning and at the end where I set up the camera that we have to get rid of first. And ideally, I want the agent to do this for me. So, I don't even have to open a video editor like Premiere Pro. So, now we can simply drag and drop these video files into our chat here to attach them. It uploads them and then we can tell the AI what we want it to do with it. We can either click these predefined buttons here, but we can also just chat with it and tell it what we want. So, in the prompt, I explained what I wanted to do. From video one, remove the first half second and the last second. From video two, use only from second three till 8 because remember our videos are a bit too long. Then create a transition between them where I teleport to the toilet after snapping my fingers. I disappear with a special effect and then I reappear on the toilet. Now it's good to be precise what exactly we want trimmed, but the AI can also do it itself. It can analyze the frames in the video and detect the correct frame where it has to make a cut. But if you already know what exactly you want to trim away, it's better to tell it right away. Let's send this. And the cool part is the video editing doesn't cost us any money because it doesn't include any LLM generations. We only have to pay credits when we actually generate something like a video or music. And now the AI gets to work. Okay, so it analyzed the video clips and it summarized the changes we wanted to make and the plan for the next step. And if we are good with that, we can say go ahead. Okay, it starts by trimming the video and it uses a tool called FFmpeg, which is a library to edit videos programmatically. And because the AI has access to this tool, it can do a lot of video editing for us. It can extract frames, it can trim videos. As you can see, it automatically analyzes the frames in the video to make sure everything looks properly. Now, in hindsight, I actually should have told the AI here to first trim the videos and then let me preview and confirm them before we start generating these animations because if the trim is not correct, then we will pay credits unnecessarily. It's better to go step by step instead of doing multiple different steps at once. But let's see the result. Maybe it works at the first attempt. And by the way, if you don't like coffee, you should try out yerba mate. It's delicious and it also has caffeine. And here we can see that the AI can actually see what's going on in these frames because it detected that I'm sitting in a gaming chair with my hand raised and I have a Pikachu shirt. And in the second frame, which is the start frame of the second video, I'm on the toilet relaxed, blah blah blah. Now, actually, while this is in progress, I want to start a second agent. I want to use the AI Hollywood agent again to prepare the second clip because I want to stitch two different special effects together. For the second video, I only use clip two because I want to add a special effect to the end of this video. Here I wrote trim the clip to keep only second 3 to8 again. At the end, make my head explode with a confetti effect. In the final frame, my head is gone, but don't add any gore or blood. Keep it friendly. So, we can run this agent in parallel with the other agent. And our first video is already finished. Let's take a look. Ah, something is wrong with my stomach. Give me a second. The port animation. Much better. Perfect. And now I want to make my head explode for a joke because I have a very simple humor. But wasn't this amazing? It worked at the first attempt. It cut the video to the correct size. It added this transition. It stitched everything together. This was almost no work. I basically only gave it one single prompt. Okay, let's finish the second clip. The AI agent already trimmed the video clip as I told it. And now it's generating the confetti effect. There it is. And then it will use Kling to actually turn this into a video. Okay, so this will be the final frame. And now it asks us to proceed and generate the video. But I actually want to make a little change because if this is the final frame, the animation doesn't really finish, right? And by the way, I use an AI voice dictation tool, Wispr Flow, to give the AI instructions. This looks good, but in the final frame, the confetti should already have settled. Also, make my neck visible like a mannequin. Again, keep it friendly. If you want to use Wispr Flow yourself, check out my review and tutorial video. I will put a link here into the top right corner. Let's send this. Okay, so this is the new final frame and I think this looks pretty good. The confetti has settled. You can see my neck here without any blood because we don't want to make this too brutal. Let's go ahead. Meanwhile, I'm going to download our finished first video. You can also expand these workflow cards here to see the details to see the exact prompt that the agent sends to the LLM. Like this is the Kling prompt and it's autogenerated for us. Okay, the AI even automatically speed up the video to make sure it's not too slow. Now, it combines the generated audio and the video. It stitches everything together. It does all of this automatically. And by the way, if you like an agent, you can like it up here. And then it is always in your likes list on the left, so you never forget it. And the agent is done. Let's take a look at the result. Much better. Yeah, perfect. That's exactly how I imagined it. And now I want to stitch video one and two together so that we have one long video. So I saved our generated video one here in this folder earlier. Remember I just attached this to this chat. So I told it I attached video one add it before our new generated clip and stitch them together. Video two starts where video one ends, but you have to find the matching frames to make sure the transition is seamless. And this is one of my favorite things about these agents. They can find the correct frames automatically so that you don't have to trim these videos manually. Again, the AI does this using the FFmpeg tool. It analyzes the frames. It finds the matching ones and then it will stitch the two videos together. Looks like the AI found the matching frames between the video clips. And here is the final result. Let's check it out. Ah, something is wrong with my stomach. Give me a second. Much better. Much better. That's a mistake. It's almost right, but there is a little part in the video that's duplicated because it didn't trim the video to the correct length. So, let's try at 10 seconds. There are about 2 seconds that are duplicated because you didn't trim the video to the correct length. Try again by removing another 2 seconds from the start of video two and analyze the frames to make sure they match. Let's give this a try. Now, it looks like it got it right. This is the final result. Ah, something is wrong with my stomach. Give me a second. Ah, much better. There's still a tiny little glitch between the two scenes, and we could prompt the AI to fix this by adding or removing another frame, but I think you get the gist, and the result is really good and funny. Now, how can we use this technology to actually make money and not just have fun with it? Obviously, if you are a video editor, you can use this to increase the quality of your videos and charge more money. And in the future, skills like this will be required because if special effects become so easy to make, you should know how to do it to stay competitive. If you are a content creator yourself, you can easily create these effects and add them to your own videos to make them more engaging and more funny, right? In the past, this was really difficult. Now, everyone can do it. And of course, if you have a particularly funny idea, you could also create a video that goes viral just because how funny it is. You can also, of course, create 9:16 videos for Tik Tok, for example. And this is really addictive. So, let's try out another agent. Infinite Kling 2.5 is also really cool because it allows you to create AI generated videos that are as long as you want, multiple minutes instead of just seconds by creating multiple clips with matching start and end frames and then stitching them together. And again, the agent does all of this automatically. Let's try this out. And this even includes an image generator. So we can tell it generate the following start frame. Pikachu in New York City, camera behind it, photo realistic, use different image models, and let me pick the best one before we generate the video. So, we can decide if we want to go step by step or if we want to let the agent do everything autonomously. But I like to maintain some control. And now it generates the same image with image gen, Nano Banana, Seedream, and Wan. And we can pick the one that we like the most. And in the process, it also enhances our prompt. So it takes our very simple prompt and creates a more optimized prompt. Photoreistic photograph of Pikachu 8K detail cinematic depth blah blah blah. And another great detail is that these images are generated in parallel. So it doesn't have to wait for the first one to finish. It generates all of them in parallel which saves us time. Now we have these different options available. Nano Banana, Seedream, and Wan. I don't like a v one because I want the camera to be behind Pikachu, right? And not in front of it. I think I like Seedream the most because I want Pikachu to run through the city. So even though Nano Banana looks nice because he's standing here on this roof or balcony, we can't really make him start running, right? So let's go with Seedream v4. We can actually click this little quote button here next to it to put it in the chat and let the AI know which one we want to pick. and I add it. Use the start frame and make Pikachu run through the city. Keep the camera locked on Pikachu and don't zoom out. Keep the shot close to Pikachu. This way we avoid that the camera drifts away too far which makes it harder to generate the next 10 seconds and extend the video. Let's send this again. It generates an optimized prompt and sends it to Kling. And first it gives us the opportunity to confirm this prompt. I don't want to make any changes to be honest. I didn't even read it. generate video with this prompt and now it generates the video. And if there's an error in the process, which sometimes happens, for example, because an API is not available or because an API blocks a certain prompt for trademark infringement, for example, the agent is smart enough to automatically retry the request or even try a different API, a different model depending on the exact error message. So these agents are really durable and I haven't run into a situation yet where I got stuck. The agent always figured out the next steps. And here's our first video clip. Let's check it out. We don't have any sound yet. We will add this later. For now, let's just get the videos right. And the quality is amazing. Look at the lighting. Okay, so this clip is 10 seconds long. But the point of this agent is to create multiple clips and then stitch them together into one long video. Let's try this prompt. Create the next clip. Make Pikachu turn to the left. Cross the street. almost getting hit by cars but barely dodging them. You can also make him jump. Heads up, the results don't always look perfect because Kling has some problems with creating unrealistic transitions. So sometimes you have to try again if you're not happy with the result again. You can just tell the agent to redo something. So now the agent extracts the last frame of the first video because this will be the start frame of the next clip. Then it gives us another prompt to confirm. Don't even read it. YOLO generate scene 2. And we can generate as many of these scenes as we want. The second clip is ready. Let's check it out. And I think this actually looks pretty damn amazing. Now we can go ahead and tell the AI what we want to happen in the next scene. Maybe let's create one more, then stitch them together. Let's try a more difficult prompt. In scene three, make Pikachu jump really high and break through the window of a skyscraper. While he's jumping, add some lightning effects. Okay, this one looks weird because the final frame has Pikachu with his face on the back. So, we have to try this again. Regenerate scene three. Sometimes you just have to try the same generation again if you don't like it. Keep in mind that these models keep getting better and better, so in the future they will make fewer mistakes. Okay, so our scene 3 is ready. It's not perfect, but it's good enough. Before we check it out, let's stitch everything together and generate some music overlay. Looks good. Stitch everything together and generate and add some fitting music and parenthesis action theme. So action music would be fitting. I think the agent uses ElevenLabs to generate the music. And it came up with the prompt itself. Here's our final video. It's 30 seconds long, but we could make this even longer. Many minutes or even hours if we are willing to spend the credits. Let's check this out. Yeah, the third scene looks a bit weird, but I didn't want to spend any more credits on this. We could go ahead and change the scene and regenerate it with different prompts, but you get the gist, right? You could use this to generate fully blown movies or long videos. You just have to prompt each scene separately until you get it right. And you could post this on social media. And what I really like about these AI agents is that they take care of the technical part. You don't need to know how to edit videos. You just have to bring in your own creativity. That's the human part. If you're not creative, your result will not be good. Let's take a look at some more examples of generations I created earlier. Another cool agent is miniature history which creates these viral history videos with this cool effect where everything looks like a toy. But I don't want to run this again. I have prepared some videos. I made one of these videos about Apple and Steve Jobs. Let's take a look at it. So I only gave the AI a topic. I told it to create a short video about the Apple company in relation to Steve Jobs. How he founded it, how he was fired temporarily, and how he came back to save the company from bankruptcy. The agent has access to Perplexity, the AI internet search. So, it automatically searches the internet for the information on this topic. Then, it comes up with a script all automatically. Then, I can confirm the script. Then, it generates a bunch of different frames for the different scenes. If I want any edits, I can tell it here. And then it goes ahead and creates different video clips from these images. It creates fitting music and voice over and captions, everything. I just gave it this one prompt in the beginning. And this video here at the bottom is the final result. Let's check it out. In 1976, two young dream and I really love this style named Steve Jobs and Steve Wozniak built the first Apple computer in a California garage. By 1984, Jobs unveiled the revolutionary Macintosh. Apple had become a billion-dollar empire and Jobs was its visionary king. But in 1985, a bitter power struggle ended with Jobs being fired from the very company he created. Yeah, again, the AI is not perfect. There are still some weird glitches like Steve Jobs here walking right through the window. Again, if we are not satisfied, we have to regenerate this again. But I think it doesn't have to be perfect. From the very company he created. 12 years later, Jobs returned to a dying apple. Yeah, another weird glitch, but again, it's fine. He launched the iMac and transformed it into the world's most valuable company. Looks super cool. Again, you can use these videos on social media to grow your followers, to create awareness around your brand. You just have to be creative. You can make these videos about any topic. And of course, you can also generate them in 9:16 ratio for Tik Tok or YouTube shorts. Another amazing agent is Slide Guru. Again, I will put the links to all these agents into the video description. This one can create presentations about any topic with cool transition animations. Let's check this out. So, I copy this JSON here. Then I get a link to a player somewhere here in the chat. The Glif slides player. Let's open this here. We paste this JSON. I think in the future this will all be handled automatically. Right now, we have to do this weird process where we paste the JSON. Then we can start the presentation. Again, this one is about the matrix plot. This is the topic I gave the agent. And now between all of these, I think we even have voice over. Let's check it out. The Matrix. What if everything you believe to be real was a lie? Then I pressed an right arrow to move to the next slide with an animation. And again, a voice over. Thomas Anderson. By day, a software programmer. By night, a hacker known as Neo. He Yeah, I skipped too early. A mysterious man named Morpheus offers. The transitions are not perfect, but I still think it's pretty cool. And you can use this to create a presentation about anything and just make it a bit better looking. Right, we have a few more slides. I want to skip this here. You have agents for different viral video trends. For example, this talking food video about cholesterol I created. Hey, I'm Steak. For decades, they told you I'd claw your arteries and give you heart disease. But here's what actually happened. In 1967, the sugar industry paid Harvard scientists $50,000 to publish a study blaming fat and protecting sugar. No disclosure, no ethics, just corruption. Meanwhile, your body literally needs my cholesterol to build hormones, brain cells, and every cell membrane you have. 50 years of lies. I was framed. This is a true story, by the way, and you should eat meat if you want to be healthy. There is a short form video agent that can create pretty much any kind of video. I created this video about this lion attack. At first, he seemed content to observe from a distance. But something changed. The guide's voice dropped to a whisper. Again, these videos can have any length by stitching multiple videos together. Nobody move. But the lion had already decided. Oh no, the lion attacks. In that final second, instinct took over for both predator and prey. Yeah, I think it's pretty cool. And this has the potential to go viral on Tik Tok. Maybe not this particular video, but what you can do with this. Now, one of my favorites is this agent that can create this exploded view animation. So, I gave it two input images. this Game Boy and the single pieces of the Game Boy. And then it created this animation. Let's check it out. Exploding view. And you can use this for product marketing. Okay, there's a little glitch in here. I just noticed it's not perfect, but we would have to try this again to get it perfect. Still, I think this is pretty cool. And you can do some interesting stuff with this. The fashion shoot agent is also pretty cool. I could show you different agents for hours because of all these cool things you can do with them. So, fashion shoot creates these different zenes and I only gave it one single input image. This picture of Leon from the Resident Evil franchise. Then the AI created these different images here and then it adds transitions between them. I also told the AI to add a backflip between the first and the second frame and this is the result we got. Let's check it out. Yeah, he does the best. Yeah, this looks super professional and high quality and all from a single input image. Let's take a look at one more. Um, this AI Tool Corner neon design I showed you in the intro is also very good. So, this is an agent particularly for creating motion graphics. Again, the links to all the agents are in the video description. And this is the result. It's so good. I love it. And I will use this in the future for my own material. We have a ton of other agents. We can also create influencers, product influencers. This is what I used for the intro. There are also different prompt helpers that just help you create optimized prompts for other AI tools like this Zuno prompter or I think there's one for Veo 3. Yeah, this master prompter that helps you generate optimized prompts. You can check out all these different agents and the library keeps growing and you can even create your own agents either just for yourself or to provide them to the community. Let's take a look at workflows too. So workflows are very similar to agents, but they are simpler because an agent can use different tools and workflows and decide when to pick which tool. Whereas a workflow is a very deterministic step-by-step process. You have, for example, a prompt. You can have different AIs in the process, but it's always a very specific step-by-step process. Let's take a look at an example that I used earlier. This one here, control net. It's simple, but pretty cool. So you can create an image that has text on it, but the text is part of the image. So for example, this image here says AI tool corner as you can see, but when we actually expand it, you can see that those are just different servers and cables, and you can't even read the text when you have it in a big viewer, it looks like a normal photo, but when we make it smaller, it says AI tool corner. It's just a cool little thing to try out. You have different meme generators for different popular memes. you have a thumbnail maker, but I think for thumbnail generations there are better tools. I have a full playlist that I will put in the top right corner with the best AI thumbnail generators. But still, it's cool to have this here. And the results are actually not bad and not bad at all. Those are some examples down here. You also have access to different models directly. For example, image models like Nano Banana Pro. And they are also specialized for different tasks with a custom system prompt. So you have this Nano Banana Pro for editing. You have another one for image generation. You have all these different optimized AI models. You have audio models like fish and ElevenLabs. Different video models of course everything. And with workflows you can even see the source. So the exact step-by-step process when we click on this view source button which shows you all the prompts and models and steps that are included in this workflow. We can even edit it. I could add additional steps. I could change the prompt or a model that's used here and then I can click on the remix button to save it again either for myself or for the community. You can also build your own agents that can use different tools and workflows. This is a bit more complicated. If you want a dedicated tutorial on creating your own agents or workflows, let me know in the comments below. Then I will make another video. But this is beyond the scope of this tutorial. You can also open workflows and agents in different other tools like chat GBT cloud Discord or access them via an MCP server. There's also a browser extension that you can check out. But I want to take a look at the pricing now. Down here you can see my remaining credits. When I started recording this video, I had 1361 credits. Let's calculate how many dollars this converts to. So when I click on credits, I can purchase more credits. The cool thing is they don't even offer a subscription. You just buy credits directly and you can access all these different tools and LLMs. I think this is pretty cool because it's annoying to have like 10 different subscriptions if you only need a certain amount of them. So, I used around 550 credits. Let's see how many dollars this translate to. Yeah, it should be around $5.50 for all the different things we did in this video. We created this infinite Pikachu video. We created a special effects video and some smaller things. This cost us five bucks. I think that's a fair price because after all, all these AI tools cost money and you don't even need a video editing tool, which is a big plus. Now, I wish they had an affiliate program because I really love this tool and I will keep using it in the future, but I can't earn any money on this video because they don't have an affiliate program. Glif team, if you're watching this, please hit me up with an affiliate deal. Again, I will put links to all the agents and workflows I used in this video into the video description yourself if you want to check them out yourself. And let me know in the comments what's your favorite agent or workflow. And subscribe to the channel for more AI tool reviews. And then I hope I see you in the next video. Have a nice day. Take care.