Today it is clear that we cannot trust anything we see or hear, but we are only taking the first steps when it comes to creating content with AI, the world of fake news and deepfakes is alone starting.
A test is what they can do at www.fal.ai/camera, an online camera in which we only have to put, in the text field, the person we want to look like, and it will begin to generate video in real time while We move our heads.
The company responsible for this solution is specialized in Real-Time Stable Diffusion APIs, which use Latent Consistency Models (LCM). Recently, fal has released APIs that take advantage of LCMs to generate images in the blink of an eye, specifically in an impressive 150 milliseconds. This is a big jump from the usual 10 seconds of traditional stable diffusion.
The speed of these APIs has already allowed the creation of demo applications such as fal.ai/dynamic, or the camera you see above. In collaboration with illumine AI, they have developed the LCM-Painter space on HuggingFace, which has achieved great popularity. Additionally, we have experimented with implementations in design tools such as Figma and tldraw, demonstrating the versatility of this technology.
Access to these APIs is done through a REST API interface, notable for its simplicity and effectiveness. It is based on POST requests with data in JSON format. This method makes it easy for developers to integrate imaging into their applications, opening up a range of creative possibilities. For even easier integration, fal offers a JavaScript client. This client greatly simplifies interaction with API endpoints, allowing developers to focus on creativity and innovation.
To achieve greater efficiency, they have also explored the use of WebSockets, a technology that allows real-time bidirectional communication between the client and the server. This can significantly increase the imaging speed, reaching between 3 and 5 frames per second.
We are faced with tools that have the potential to change the way we interact with technology in our daily lives. Imagine artists creating works in real time, designers modifying prototypes instantly, or educators using generated images to illustrate complex concepts. We are facing a significant change in digital creation.
