Generated by Pika

Pika Labs is an innovative startup specializing in developing an AI-powered platform for editing and generating films using captions and still images. Founded in 2023 by Demi Guo and Chenlin Meng, the company revolutionizes the video creation and editing process by integrating cutting-edge AI technology.

The platform offers user-friendly tools for everyone. One of the remarkable features is Pika Art, which takes text and/or image prompts to generate mesmerizing short videos. By using a text prompt, we created the head image of this article. We asked it to create a flow diagram, and it did present a flow diagram, in a literal way.

On the contrary, the following flow diagram is manually drawn, illustrating on how Pika Art works:

Image by author

Pika Art is completely free to use, although there may be a waiting queue for access. However, it is definitely worth the wait to experience the latest advancement in AI-driven text-to-video generation. To join the waitlist to explore Pika Art, visit https://pika.art/waitlist.

In this article, we provide guidance on how to use Pika Art, via the Pika Labs Discord server and the Pika Art website.

Part 1: Discord Server

There is getting-started channel on the Pika Discord server, describing how to use the service.

Image by author

It introduces 10 channels to use, with no differences among them. Here is a snapshot of one of the channels, displaying user generated 3-second videos:

https://youtu.be/ZP68igSrG_s (video by author)

Here is the command list on the Pika Discord server:

• create
• animate
• encrypt_text
• encrypt_image

Let’s see how they work.

create

The create command is defined as following:

/create prompt +1 optional

prompt refers to a text prompt that specifies an instruction to generate a video. It typically consists of a few sentences that sets the context and provides initial input for the AI model to generate a short video.

Input the prompt: A lion is chasing a rabbit in the forest.

Image by author

It generates the following video:

Generated by Pika

It is sort of correct, except that the lion is not really targeting the rabbit. We adjust the prompt to be: A lion is chasing a rabbit to eat him in the forest.

Generated by Pika

The new video above is more accurate.

The style can also be adjusted by the prompt. We can see the effect with the new prompt: A lion is chasing a rabbit to eat him in the forest in Pablo Picasso style.

Generated by Pika

The text prompt can include scene description, action description, medium and style specification, camera position and settings, along with parameters.

Here are some handy parameters:

• -ar: It sets aspect ratio of width and height. The valid ratios are 16:9, 9:16, 1:1, and 4:5. The default ratio is 16:9 (1280 x 720). Here is an example: -ar 16:9.
• -fps: It sets frames per second. The range is 8–24. A higher number keeps the motion smooth with crisp details. The default value is 24. Here is an example: -fps 24.
• -motion: It adjusts the strength of motion. The range is 0–4, and the default value is 1. Here is an example: -motion 1.
• -gs: It specifies the guidance scale for how much the result should be related to the text prompt. The range is 5–25, and the default value is 12. Here is an example: -gs 12.
• -neg: It specifies the negative prompt on what should be excluded from the result. The value is a list of unwanted strings. Here is an example: -neg "ugly, deformed, unclear, distorted, blur".

Try a text prompt with some parameters, such as : A lion is chasing a rabbit to eat him in the forest in a Pablo Picasso style. -ar 9:16 -motion 4 -neg "ugly, deformed, unclear, distorted, blur". And here is the generated video:

Generated by Pika

The create command has +1 optional to add an image prompt.

Image by author

Type the prompt: A garden is in a heavy rain Image, along with the following image prompt:

Image by author

It generates the following video:

Generated by Pika

animate

The animate command is defined as following:

/animate image +1 optional

It can take an image prompt without the text prompt.

Image by author

Here is the input image:

Image by author

And here is the generated video:

Generated by Pika

It is indeed a nice addition to add some vegetable to the dish.

encrypt_text

The encrypt_text command puts in a message in the video as a visual element rather than uses it as part of the prompt. It needs a message, a text prompt, and 2 optional choices of font and image.

/encrypt_text message prompt +2 more

Image by author

Use the message, A summer day, and prompt Text is moving in the blue, along with the following image prompt:

Image by author

Here is the generated video:

Generated by Pika

In the above image, we have picked the font Modern, among the 5 available fonts, Modern, Comics, Sans Serif, Bauhaus, and Retro.

Image by author

There are also 2 additional parameters:

• -w: It stands for weight on how much attention to pay to the uploaded start image. The range is 0–2. The default weight is 1.
• -size: It sets the size of the font. The range is 50–100. The default size is 100.

encrypt_image

The encrypt_image command takes the encrypt image as a mask to draw with. The command has a message image, a text prompt, and 2 optional choices of font and image.

/encrypt_image message prompt +2 more

Image by author

We use the following encrypt image, along with the prompt: The image flows on the water surface -w 2 -motion 4.

Image by author

Here is the generated video:

Generated by Pika

With the same setting, we add a start image:

Image by author

Here is the generated video:

Generated by Pika

Adjust the prompt to remove the lady in the sky: The image flows in the sky -w 2 -motion 4 -neg ''people''.

Generated by Pika

The video above complies with the instruction of the mask R, the prompt, and the start image.

Part 2: Pika Art Website

Pika Labs also provides a website for Pika Art:

Image by author

It has two tabs, Explore and My library. Explore showcases featured videos, and My library contains all user generated videos.

Here is the generated video, with the the prompt: angel flying over mountains, dark fantasy, insanely detailed, 4k.

Generated by Pika

By default, there is no camera control.

Image by author

Add some camera controls, with the same prompt.

Image by author

This is the re-generated video with camera controls:

Generated by Pika

The following is another generated video, with the the prompt: a spinning spaceship that travels through milk way, 4k.

Generated by Pika

The generated image has a choice to Upscale.

Image by author

Can you tell that the following upscaled video has higher resolution?

Generated by Pika

Here is a generated video, with the prompt: slow overhead shot over luxury mansion surrounded by forest, high fantasy, dynamic movement, 4k.

Generated by Pika

The generated image has a choice to Add 4s.

Image by author

A newly generated video is 7-second long.

Generated by Pika

We can Reprompt the video: a dog and a cat dance in front of the mansion. The original video becomes the video input.

Image by author

Here is the new video with pets jumping around in front of the mansion.

Generated by Pika

By trying out Pika Art, we are building up our own library:

Image by author

Conclusion

Pika Art represents an innovative leap in AI-powered text-to-video generation. Our comprehensive guide on utilizing Pika Art is readily available on both the Discord server and the Pika Art website, accompanied by illustrative examples.

Are you going to give it a try?

Thanks for reading!

Want to Connect?

If you are interested, check out my directory of web development articles.