Venturing into the world of AI-powered video generators can be quite the labyrinth, filled with promises of innovation and efficiency. However, it's not always easy to find the right tool that suits your unique needs – and this goes for any generative AI. So, we dug a little, and tested various video genereators. Now, bear in mind that we were looking for creative AI video generators, so any corporate video generators – like Synthesia –, are out of this list.
Now, without further ado, let's dive into the five picks of the moment.
This is, for the moment, the most useful tool for a filmmaker. It has many “magic AI tools”, but we are going to talk only about Gen-1 and Gen-2.
Gen-1 lets you upload your video and apply any desired style to it. This can be achieved by uploading an image that encapsulates the specific aesthetic you're aiming for, but do make sure this image has some similarity to the composition of your video (if not, you'll have a monstrosity!). Alternatively, you could use one of the tool's presets (metal, cloudscape, etc.), or describe the style you’re looking for through a prompt.
Gen-2 is a text-to-video generator. The process is straightforward: write your prompt (for instance, "astronaut in an alien jungle") and wait for the results. Now, it’s still in its baby steps and lacks the ability to create John Wick kind of action scenes. Ok, we'll be honest... even the execution of simple action can prove challenging, but it does have this weird AI-looking vibe that provides an opportunity to play with.
Why we like it? It has a video-feel to it, it’s somewhat fluid, does not flicker and is easy to use.
The downside? The price. It is $15/month (standard), which gives you 625 monthly credits. Each generation costs, so it’s not the most affordable tool when it comes to iteration:
1 second of Gen-1 generation = 14 credits ($.14)
1 second of Gen-2 generation = 5 credits ($.05)
This an incredible tool, the full potential of which we've yet to exhaust, but we are already amazed by the results. Available for free download, Ebsynth operates on a principle somewhat akin to Gen-1, but it’s a little bit more difficult to use.
Their concept is that you can “transform a video by painting over a single frame”. Well, it’s more over a few keyframes, but if you do your work properly, you’ll have neat results. This approach is ideal for filmmakers keen on collaborating with a specific artist and vice versa.
How does it work? First, you need to convert your video into an image sequence and generate an alpha channel. Next, select keyframes for the artist to meticulously paint over. The final stage involves rendering your creation through the tool.
Why we like it? It is very faithful to the video and the artist’s style. It is free.
The downside? You must have the final outcome in mind when filming, because it does not generate “new” things over your video. For example, you film a pirate but forgot the hat…, well Ebsynth won’t let you add it, even if you paint it.
For this test, we changed the style of one keyframe with Stable Diffusion (as we're not that kind of artist) and then rendered it through Ebsynth.
Kaiber is a tool “made for artists, by artists”. This video generator has an user-friendly interface. It offers multiple options: one can input a prompt, upload an image to be transformed into a video, or even alter the style of an existing video. It also provides a range of adjustable parameters such as aspect ratio, camera movement types and a variety of preset styles. It makes some cool stuff, but personally, it is not the AI that we preferred.
Why we like it? It is fun and easy to use.
The downside? Inconsistencies in output and noticeable flickering between each image. In our testing, the most successful results were achieved when we uploaded a video and applied a distinct style to it (see below). Starting price: $5 / month for 300 credits.
Test made using an input video and rendering through Kaiber
You've probably encountered countless videos created with this tool on TikTok, as it rode the wave of a trend on the platform, but it’s a headache (at least for us). First and foremost, this is an extension of Stable Diffusion. Deforum can be used to create animation videos by providing the prompts and settings for how the camera moves. It uses Stable Diffusion's image-to-image function to generate a series of images and stitches them together to create a video.
This tool is free, and you can run it on various platforms such as Google Colab, Windows, and Mac. Depending on the platform you are using, the installation process may vary. However, it isn’t the most user-friendly tool and you need to have a lot of patience with it. There are many settings you can play with (or get lost with), you need to learn how to write a Deforum prompt (which is not a simple sentence – it has its science) and a high-performance computer for smooth operation.
Why we like it? It's free, with outcomes that can surprise you. You can also use a video or image as a base.
The downside? A complicated tool with many things you have to figure out. Lot of flickering.
Promo for Infidelitas, made with Deforum
This text-to-video AI tool is available for everyone to use! It is a "Modelscope-based video model optimized for producing high-quality 16:9 compositions and a smooth video output".
It is totally free and you can try to prompt some of your ideas! Mind you, it takes time to render (sometimes it is even saturated - like ChatGPT at the beginning) and it's not super high quality – as promised -, but it is a promising tool!
Why we like it? A useful and fun tool to try out a video generator.
The downside? If used as a free tool, it can be very slow. It is difficult to render a consistent idea, as you cannot put a “seed”.
Test made using similar prompts to the ones used with Gen-2
For those who have stayed with us this far into the article, here's a pro-tip:
If your goal is to create an AI video or incorporate AI elements into your videos, we strongly recommend experimenting with various tools. Find the one that aligns with your needs. Consider blending different tools for better results. Some creators have repurposed D-iD, originally designed for corporate videos, into a tool for crafting fictional short stories (we teted it too!). Most of the tools have a free trial… our advice? Try them all. You’ll quickly see which one fits best.
コメント