Kandinsky AI is a web-based AI video and image generation platform that produces high-resolution clips from text prompts or uploaded images, using a family of diffusion transformer models developed by Kandinsky Lab.
What is Kandinsky AI?
Kandinsky AI is a text-to-video, image-to-video, and text-to-image generation platform built on a latent diffusion pipeline trained with the Flow Matching paradigm and a scalable Diffusion Transformer (CrossDiT) backbone. Users input a text description (up to 2000 characters) or upload an image (PNG, JPEG, WEBP, BMP, TIFF, GIF; max 30MB, dimensions 300–6000px, aspect ratio 0.4–2.5) and receive a video up to 10 seconds long or a high-resolution image. The platform runs entirely in the browser and is offered as both an open-source model family and a paid web service.
Key Features
- Text-to-Video Generation — Convert plain text prompts into 5‑second or 10‑second video clips with smooth motion and coherent storytelling; choose between 480p, 720p, or 1080p export depending on plan.
- Image-to-Video Animation — Upload a static image (supports multiple formats) and animate it while preserving structure, style, and subject identity; optional first‑and‑last‑frame control.
- Text‑to‑Image Generation — Generate high‑resolution, detailed images from text descriptions in seconds, suitable for marketing assets and digital art.
- Style Consistency & Concept Control — Maintain brand identity and character continuity across multiple video and image generations; models include SFT, Flash (16‑step distilled), and RL‑finetuned variants for different quality/speed trade‑offs.
- Model Family Options — Choose from Video Pro (SFT model for highest quality) or Video Lite (faster generation) modes; image generation offers SFT‑soup and RL‑finetuned checkpoints.
- Commercial Use License — All paid plans include a commercial use license; videos export without watermark at 720p or 1080p depending on tier.
Who is it for?
- Marketing and advertising teams — Create brand‑consistent video ads and promotional imagery without a production studio.
- Social media content creators — Produce 10‑second AI videos and images optimized for TikTok, Instagram, YouTube Shorts, and other short‑form platforms.
- Digital artists and illustrators — Generate high‑fidelity concept art and illustrations from detailed text prompts, then refine with image‑to‑video features.
- Storyboard and animation designers — Turn scripts or rough image concepts into animated clips for pre‑visualization and motion design prototyping.
What can you do with Kandinsky AI?
- Campaign and ad creation: Generate cinematic 10‑second clips (e.g., a train chase or bank heist scene) for use in broadcast or digital ads, with consistent lighting and camera movement.
- Social media content: Quickly produce looping 5‑second or 10‑second videos for product showcases, teasers, or background visuals — export in common aspect ratios (16:9, 9:16, 1:1).
- Storyboarding: Animate static storyboard panels into short motion clips to communicate scene pacing and camera angles before full production.
How does Kandinsky AI work?
- Write a text prompt (or upload an image) describing the desired scene, style, and motion.
- Select output type — text‑to‑video, image‑to‑video, or image generation — and set duration (5 or 10 seconds) and resolution.
- Click “Generate Video” – the model processes the input using the chosen pipeline (e.g., Flash for speed, SFT for quality) and returns a video clip.
- Download the result in standard formats (MP4, etc.) and integrate into your workflow.
