[0:00]I've spent the last three years creating AI images and videos every single day. And what I learned is this. The best AI videos aren't made with complicated prompts. They're made with simple intentional techniques that give you control and save you money. So today, I'll show you six skills that turn AI video generation from trial and error into a repeatable workflow you can trust. The first skill is choosing the right AI video approach. There are two methods and each situation demands a different input. When consistency, the second method becomes crucial. But first, let's start with the first method, perfect for quick social videos. When for example, you'll add your logo and branding later in editing. And I will do this on Higgsfield AI, an all-in-one platform where I have all my favorite AI image and video models under one subscription. And with the Kng 2.6 video model selected, I will use the text to video option and I added my cinematic product promo prompt in here. Now I'm going to select audio, because I want to have a voice in it, then I'm going to set the duration to 10 seconds and then hit generate. And that created this video.
[1:13]Level up your AI skills in 2026. Text to video is super useful when you need a usable clip fast. But the downside is, you have less control, doing more runs and spending more money. So if you want full control on your generations, I recommend the second method. And this is image to video. Here you split the job into two steps. First you create the image. I will dive deep into how to best create them in the next skill. Then you use this as an input with an image to video prompt like this. And the cool thing is, now you only have to focus on the camera move and the action. So I used the camera zooms into the woman's face for the camera move and for the action, I used, and she does not attacks, she simply relaxes out of the low training stance and she slowly powers the saber down. Now to be able to use image to video, you just have to upload your image in here and you can see I've added my prompt in here and I've still have the clean 2.6 models selected. And then this is the final video. And because images are way cheaper than videos, you can tweak them until they look exactly right, without wasting too much credits. And it's also much easier to keep your characters and style consistent with images, which I'll show you later in this video. The second skill is picking the right text to image prompt method. Because the wrong one will leave you tweaking forever and slows down your workflow. That's why I'm going to show you two methods. Each one works best for its own situation. Method one is using LLM's like JGPT or Gemini. Nano Banana Pro is one of the best AI image models right now. And if you want an LLM to write prompts for it, I recommend Gemini, because Nano Banana Pro is powered by Gemini, so it speaks the same language and knows what the model responds best to. But to be on the safe side, I created the Nano Banana Pro Writer Lock Prompt. So I pasted that prompt in here and then I asked for a Jedi woman and then I got this prompt. So now I'm going to exactly copy this and in the image tab with the Nano Banana Pro model selected, I've pasted my prompt in here and I've set the aspect ratio to 16x9 and the resolution to 2K. And that created these two amazing images. But here's the catch. LLM's usually spit out long prompts, and tweaking them is slow and annoying. The truth is, shorter prompts give you more control and a faster workflow. Method two, short prompt structure. Now, here's what I mean. Start with just four simple components. That way, you can instantly see what each word is doing in the image. So, for the subject, I took a Jedi woman, for the composition, I used a full body shot, and for the action, I used in a focused stance. She holds an ignited cyan lightsaber with both hands in a 45° angle, and for the location, I used in a rocky vast desert. And that gave us these two amazing images. Now, to make this image more cinematic, you can add some style keywords, like for example, lit by low angle golden sunlight with long cinematic shadows and dust glowing in the air. And that gave us these two images and in my opinion, they look very cinematic. Now, let's go one step further and create a consistent world from shot to shot. The third skill is character and world consistency. Because without it, you're not building a story, you're rolling the dice every shot. And your images will feel random. But when you prompt the details on purpose and repeat them as your base, everything feels like it belongs in the same world. There are two pillars to consistency and pillar one is scenery. The foundation your whole story depends on. Now, you could upload the scenery image into Nano Banana Pro as a reference, and this works really well, but here's the catch. If that image has a golden hour of sunset look, that exact vibe will get copied into every new shot. That's why it's better to build your own consistent world using anchor keywords, keywords with a fixed identity. There are many anchor keywords you can use. Let's start with the first one, style, and for that, I used military sci-fi. And for the location, I used white underground tunnel. And with that prompt, I got these two images. The next anchor keyword you can use is mood. And for that, I used in a gloomy cinematic atmosphere. And that gave me these two images. Now, let's add another anchor keyword, lighting. For this one, I used in bi-color global illumination, environment saturated in dark orange and cerulean. And that gave me these two images. Now, there are also realism anchor keywords you can use and I'll show you those in a second. But first, look what happens when I remove the military anchor keyword. The entire scene shifts, but the cinematic style and overall look stays the same. And you'll see how these small changes give you full control over your scenery creation. And for realism, you can use camera and lenses. And I used shot on an Ari Alexa 65, F 1.2, 35mm wide angle lens. And that gave me these two images. Now, if you want a clean shot of your scenery to use it as an image reference, you can use negative keywords. So I used no people, vehicles, text, watermark, camera info, date stamp. And then you get images like this and that feel like they belong together. Now let's combine our anchor prompt with the second consistency pillar. And that's how you keep your character and outfit consistent. You can do this with one simple character sheet. And there's a full tutorial showing how it works with a single prompt on my YouTube channel. So now we'll use this as the input image. You can see the anchor prompt highlighted in green, and for the subject prompt, I added wide shot of the man who is in a low ready breach stance, holding a compact matte black rail gun with a glowing holographic sight, inside a military sci-fi white underground tunnel with bright white walls in a gloomy cinematic atmosphere in bi-color global illumination, environment saturated in dark orange and cerulean. Now, back on the image tab in Higgsfield AI with the Nano Banana Pro model selected, I uploaded my character sheet input image in here, and I added my prompt in here. This time, I will select 4K and I will hit generate. And that gave me this image. As you can see, because of the anchor keywords, you can create images that feel like they belong together. And now we can take the same character and place them in any location, while still building a clear cinematic story. And in the sixth skill, I will dive deeper into this. Now, one last step for the anchor keyword technique. If you want to have more grain in your shot, you could use a film stock like Fuji Superior 100 film. And if you want to change your look a bit, you could change for example, gloomy or cinematic, or remove the word dark, and your scenery will change slightly. Now, even with a perfect film look and consistency, your videos can still feel like they don't have your fingerprint yet. The next skill fixes that. Skill four is using cinematic consistency presets. Instead of prompting forever, you use simple sliders to lock in a Hollywood style look. It saves time, saves credits and keeps everything consistent. We'll break this down into two steps. First, we lock in a consistent look with the sliders. Then we combine it with the anchor keywords to create the ultimate cinematic look. Inside Higgsfield's Cinema Studio, we start in the image tab. It works just like Nano Banana Pro. You upload an image and build from there. The big difference is that here you can choose camera and lens settings. By mixing these, you can quickly create many different cinematic looks without heavy prompting. And one of my favorite combos is the Sony Venice camera body with the RE Signature Prime lens, with a focal length of 35mm with an F 1.4 aperture. And then I used a very simple prompt, this man as a Jedi in a Star Wars movie scene. And that gave me this cinematic looking image. And with the same camera settings, I created some more images. And they all have the same cinematic look, it all feels connected. Now, let's move over to the video tab by clicking on the animate button over here. Now I will add my prompt in here. And here I can click on the movements button and there's a lot of magic here because you can choose different camera settings. And I will choose the orbit around one. I will set the duration to 5 seconds. I will leave the audio on, because I want the Jedi to say something, and then hit generate. And then we got this amazing video. This battle is over now. Now, let's combine the anchor keywords from the last skill with Cinema Studio to get the ultimate cinematic look. And in the image tab, you can see I've uploaded my character sheet again, and I added my prompt in here. And I found that these camera settings added the final cinematic touch to my anchor keyword look. And this is the result. The change is subtle, but everything feels sharper, cleaner, and more cinematic. Now, even with a perfect film look and consistency, your videos can still feel like they don't have your fingerprint yet. The next skill fixes that. The fifth skill is creating your unique signature look. And this matters because if you keep copying famous films for inspiration, you'll end up with the same overused look, and you can even run into copyright issues. Star Wars is the perfect example. If you prompt Star Wars, you get the obvious risky results everyone else is already making. But here's the solution. Every film style is built on a genre foundation, and once you know that, you can create something that feels familiar, but looks original. So let me show you the three steps to build your own unique look. Step one, the visual foundation prompt. Let's say you want a Star Wars Stormtrooper style character. Then in step one, we ask, what is the core foundation of Star Wars? Then in step two, we ask three to five copyright safe alternative names for Stormtrooper, and in step three, we ask for a clear visual description from the outfit. So then I copied the whole prompt into Gemini and got the following answers. The foundation for Star Wars is Space Opera. And here we got some different character namings and I'm going to use Void Guard. So with that, I made the following prompt. Full body shot of a Space Opera sci-fi void guard on a gray solid background. Here I added some camera info and some negative prompts, and then I set the aspect ratio to two by three. And that gives us these kind of images. And they already look like a Stormtrooper, but different. Now, if you want your character to look more like a Stormtrooper, you can use information from step three, from the outfit description. So, that's why I've added wearing a glossy white composite armor with an integrated respirator module.
[11:40]And that gives you results like this, a Stormtrooper based character with his own signature loop. And by adding more or less details from step three, you decide how close or how far it stays from the original base. You can use the foundation prompt for characters, vehicles, and locations. Now that you know how to create your own style, now it is time to direct your shots like a film. The sixth skill is cinematic storytelling, and this is important because it's the difference between a nice AI clip and a shot that feels like a real film. There are three elements that matter. Let's start with the first one, choosing your location. And this is where the story starts. The place you choose already sets the tension, the stakes and the mood. A crater base with landing pads feels like a high value target. Something important is about to happen here. This outpost room feels empty and isolated, like they calm before the next fight. A white desert feels exposed, wide open and nowhere to hide. A spaceship hangar feels massive and dangerous, but there's loads of hiding spots. With element two, we plan the camera angle. And with element three, we add the movement. And together, that's what brings the story to life. An off-center shot instantly feels unstable, and when we slowly pan sideways, it creates anticipation, like you're searching for the threat. A close-up strips away context and forces the viewer to focus on emotion, and the slow push-in pulls us deeper into the character's head and raises the suspense. A high angle makes him feel small and watched, like the world is bigger than him. And by zooming out, we amplify isolation, like he's alone in a huge battlefield, with danger everywhere. An extreme low angle shot makes him look powerful and dominant, like he's taking control. And the fast zoom into his face, feels like you're being pulled into the moment with him. And the ground level angle makes it feel like you're right there with him, moving through danger. And the tracking shot builds suspense and tension, making it very personal. If you want more cinematic video skills, make sure to watch this video, and a big thanks to Higgsfield AI for sponsoring this video. The link is in the description.
