Kavel is an AI image and video generation platform that consolidates multiple generative models into a single interface. Its primary function is to allow users to create images and videos from text prompts or reference photos, using models such as Nano Banana 2 for image generation and Seedance 2 or Veo 3.1 for video generation. The platform displays the estimated credit cost for each generation before the user commits to running it, and failed generations are not charged. Users can sign up to receive free credits to test the service before purchasing. The main features of Kavel include image generation and video generation, both accessible from the same studio. For images, Nano Banana 2 can produce outputs ranging from 1K to 4K resolution based on text descriptions, and users can incorporate up to 14 reference photos to combine, edit, or transfer styles. For video, Seedance 2 and Veo 3.1 generate moving clips from text or images, with the ability to specify start and end frames. Veo 3.1 supports outputs of 4, 6, or 8 seconds and includes native audio in the same generation pass. The platform also supports editing from reference photos and generating video with synchronized audio in a single step. Typical use cases include turning a written idea into an image or short video clip, producing cinematic footage with controlled motion, and repurposing existing photos into new visual content. The workflow involves selecting a model suited to the specific task (images with Nano Banana 2, video with Seedance 2 or Veo 3.1), entering a prompt or uploading reference images, reviewing the estimated credit cost, and then generating the output. The interface keeps generation, editing, and credit tracking on one screen, reducing the need to switch between separate tools. Technically, Kavel operates on a credit-based system where the cost varies by output resolution, video length, and model choice. The platform supports both text-to-image and text-to-video pipelines, as well as image-to-video and reference-based editing. It is web-based and accessible via the kavel.ai website. The service is designed to let users match the model to the job rather than relying on a single generator for all tasks.