How To Create Cinematic AI Videos With Pika Labs

Bring in a Cinematic Image

So, as mentioned above, for this guide we’re going to be following a tutorial laid out by the team at Curious Refuge. Developed as the “world’s first online home for AI storytellers,” Curious Refuge is the channel behind those popular Star Wars by Wes Anderson trailers that we’ve covered before.

More notably though, Curious Refuge also offers its own AI filmmaking course which, if this is your interest and cup of tea, you can check out here.

To get started though, all you need to do is create an image. This can be something you create with AI or dream up the good old-fashioned way on your own. For this tutorial, our host recommends using Midjourney as it’s still one of the best text-to-image AI tools available.

Open Image in Pika Labs

From there you’re going to want to head to Pika Labs, which currently operates as a Discord server much in the same way as Midjourney. So, if you’re familiar with pulling AI images from prompts in Midjounrey, you can expect to use Pika Labs much in the same way.

And, if you aren’t familiar, it’s still pretty easy.

Prompt and Direct

All you’re going to need to do is enter a prompt and direct your video scene. To upload your image all you need to do is drag and drop your file into the Discord channel. From there you’re going to type in the type of action that you’d like to see.

Caleb from Curious Refuge gives some good insights into what makes a good prompt, but you’re free to try out as many different approaches as you’d like in this phase. Pika Labs is surprisingly knowledgeable about filmmaking as you’ll be able to enter prompt commands for adding camera movement, zoom or rotation, as well as even changing the frames per second.

From the Curious Refuge tutorial, Caleb recommends using these tags at the end as a way to remove any unwanted qualities: “-neg morphing, erratic fluctuation in motion, noisy, bad quality, distorted, poorly drawn, blurry, grainy, low resolution, oversaturated, lack of detail, inconsistent lighting.”

Refresh or Shuffle As Needed

From there all you’ll need to do is hit enter and wait (approximately around 30 seconds) for your AI video to render and load. You’ll then be able to review your AI clip and decide if it will work, or if you’d like to refresh it (to let the AI run the prompt again as is) or choose to shuffle (which will allow you to tweak your prompt before sending it through again).

This is where you can really explore Pika Labs and its AI powers to learn more about what types of prompts work for you and your desired cinematic video looks.

Remove the Watermark

Now, one downside to using Pika Labs — and many other AI apps for that matter — is that your video is going to have a big watermark on it in the corner. This isn’t a big deal if you’re just trying this out for the exploration of course. (And, if you do want to be ethical, if you’re just sharing with friends it wouldn’t hurt to leave this watermark on so you can let others know where it was created.)

Upgrade the Resolution and Export

Finally, to really give your video an upgrade and boost, our guide recommends taking your video into the Topaz AI app to upgrade its resolution and export as a larger file size. Topaz Labs has a pretty solid upgrade offering that allows users to turn your SD video into a higher-resolution 4K or 8K version.

You can also use a different AI model (like the Proteus AI model) to help smooth it out before exporting it. There are several export setting options as well, but ProRes 422 Standard will probably be the best option.

And from there, you’re done. You have a simple, yet fully realized, cinematic AI video which came from nothing but your imagination and the complex algorithms of artificial intelligence.