Melius is a node-based AI canvas where an agent named Mel routes prompts to image, video and audio models, shows every prompt, and lets creatives refine results. It targets agencies, marketers, filmmakers and e-commerce teams.
Melius is a browser-based creative workspace built around AI agents. Instead of opening a separate app for each image, video or audio model, a user describes the result they want and an agent called Mel picks a suitable model, runs the workflow and exposes the prompts it used. The goal is to let a person keep adjusting the output until it matches the original idea, rather than accepting a single one-shot result.
The core interface is a node-based canvas. Each node is a step such as a product mockup, a studio shot, a character study or a short video clip, and each one displays which model produced it. The homepage showcases models from several vendors, including GPT Image 2, Nano Banana Pro, Seedance 2.0, Kling 3.0 Omni, Ideogram 4, FLUX 1.1 and Veo 3.1. Chaining nodes lets one approved frame feed the next stage, for example moving from a still reference to a lifestyle video clip.
The target audience is commercial creative work. Melius organizes its pitch around agencies producing concept boards and campaign variants, filmmakers and creative directors building storyboards and lookbooks, marketers who need localized and resized ad variants, e-commerce teams producing pack shots and on-model imagery, and growth teams that want event graphics or blog headers without filing a design ticket. A long customer logo strip suggests adoption among consumer brands and small studios.
Billing is credit based. Individual plans (Creator, Growth, Professional) include a monthly pool of credits with approximate image and video equivalents, while the Team plan pools credits across seats and Enterprise is custom. Higher tiers add agent skills, a Slack agent, AI-assisted search and prompt enhancement.
Compared with single-model generators such as Midjourney, Melius positions itself as an orchestration layer: it trades the depth of one model's native interface for access to many models, a shared canvas and agent-driven routing. Teams that want to mix still and video models in one reviewable flow are the clearest fit, while users committed to a single model may find less reason to switch.