# Melius

> The creative canvas for agents. Describe what you want; Mel routes the right model, runs the workflow, and shows you every prompt — so you can steer until it lands.

Melius (also known as Melius AI) is a web-based node-based creative canvas for generative AI. You build work as a graph: each node is a creative unit (a prompt, an image, a video, an audio clip, a file) and each edge passes output from one node into the next, so a multi-step pipeline (reference image to image to video to voiceover to final cut) lives visually on one infinite canvas instead of across separate tools and chat threads. Describe what you want and the Mel agent wires the nodes for you, routes each one to the right model, and shows every prompt so you can rewire or re-run any step. It is not affiliated with Melius Research, Melius Inc, or other organizations that share the "Melius" name.

Focus areas: Node-based creative canvas, AI canvas, Node-based AI workflows, Infinite canvas for creative work, AI image generation, AI video generation, AI audio and voice generation, Generative AI, AI creative tools, Multi-model AI workflows, AI agents for creative work.

## How the node-based canvas works
- Nodes are the creative units: text prompts, images, video, audio, and uploaded files. Each node holds its own model, settings, and version history.
- Edges connect nodes and pass output forward, so one node's result becomes the next node's input. That is how multi-step pipelines are built.
- The canvas is infinite and zoomable. Branches run side by side, so you can compare variations of the same step instead of overwriting them.
- Any node can be re-run after editing its prompt or swapping its model; downstream nodes re-run from the new result.
- The Mel agent builds and edits the graph for you. It creates nodes, connects edges, picks models, and runs the workflow, and every prompt it wrote stays editable.
- Reusable templates capture a working graph so a proven workflow can be re-run with new inputs.
- Canvases are real-time multiplayer, and external agents can drive the same graph over MCP or the CLI.

## What Melius is good for
- Video production: storyboard, generate shots, keep takes on separate branches, then assemble and export in the built-in multi-track editor.
- Advertising and campaign creative: build one graph per concept, then fan a locked concept out to every placement and aspect ratio.
- E-commerce product imagery: reuse one product photo as a reference node across lifestyle scenes, backgrounds, and marketplace crops.
- Brand and design systems: pin brand assets, fonts, and reference images as nodes so every downstream generation stays on-brand.
- Marketing content at volume: templates plus batch runs turn an approved workflow into repeatable output.
- Agencies and creative teams: shared canvases make the whole pipeline reviewable, so feedback lands on a specific node rather than a final file.

## Key pages
- [Melius home](https://melius.com/): What Melius is and how the node-based creative canvas works
- [Pricing](https://melius.com/pricing): Plans, credits, and what each tier includes
- [Models](https://melius.com/models): Every AI image, video, audio, and text model that can run on a canvas node
- [Enterprise](https://melius.com/enterprise): Melius for teams — collaboration, security, and volume pricing
- [Blog](https://melius.com/blog): Guides, model breakdowns, and creative workflows

## AI models
- [Better Fonts (Melius, Image model)](https://melius.com/better-fonts): Melius' typography-tuned build of GPT Image 2 for crisp, legible text
- [GPT Image 2 (High) (OpenAI, Image model)](https://melius.com/gpt-image-2-high): OpenAI's flagship image model at its highest quality
- [GPT Image 2 (Medium) (OpenAI, Image model)](https://melius.com/gpt-image-2-medium): OpenAI's image model, balanced for speed and quality
- [GPT Image 2 (Low) (OpenAI, Image model)](https://melius.com/gpt-image-2-low): OpenAI's image model, tuned for fast, cheap iteration
- [MAI Image 2.5 (Microsoft, Image model)](https://melius.com/mai-image-2.5): Microsoft's MAI image model for generation and editing
- [Arrow 1.1 (Quiver, Image model)](https://melius.com/quiver-ai): Quiver's Arrow image model for generation and editing
- [Arrow 1.1 Max (Quiver, Image model)](https://melius.com/quiver-ai-max): Quiver's highest-quality Arrow image model
- [Ideogram 4.0 (Ideogram, Image model)](https://melius.com/ideogram-4.0): Ideogram's image model, strong at typography and design
- [Nano Banana 2 (Google, Image model)](https://melius.com/nano-banana-2): Google's fast, cinematic Nano Banana image model
- [Nano Banana 2 Lite (Google, Image model)](https://melius.com/nano-banana-lite): The fastest Nano Banana model, built for quick iteration
- [Nano Banana Pro (Google, Image model)](https://melius.com/nano-banana-pro): Google's reasoning image model for brand and text fidelity
- [Flux Schnell (Black Forest Labs, Image model)](https://melius.com/flux-schnell): Black Forest Labs' fastest FLUX image model
- [Flux Pro 1.1 Ultra (Black Forest Labs, Image model)](https://melius.com/flux-pro-ultra): FLUX Pro at ultra resolution and detail
- [Flux 2 Max (Black Forest Labs, Image model)](https://melius.com/flux-2-max): Black Forest Labs' flagship FLUX.2 image model
- [Seedream 4.5 (ByteDance, Image model)](https://melius.com/seedream-4.5): ByteDance's Seedream model for image editing
- [Seedream 5.0 Lite (ByteDance, Image model)](https://melius.com/seedream-5-lite): The fast, lightweight Seedream 5.0 editing model
- [Seedream 5.0 Pro (ByteDance, Image model)](https://melius.com/seedream-5-pro): ByteDance's flagship Seedream 5.0 image model
- [Uni-1.1 Max (Luma, Image model)](https://melius.com/uni-1.1-max): Luma's Uni image model with reference-image consistency
- [Krea 2 Large (Krea, Image model)](https://melius.com/krea-2-large): Krea's high-quality Krea 2 image model
- [Krea 2 Medium (Krea, Image model)](https://melius.com/krea-2-medium): Krea's faster Krea 2 image model
- [Grok Imagine Image 2.0 (xAI, Image model)](https://melius.com/grok-imagine-2.0): xAI's latest Grok image model for generation and editing
- [Grok Imagine (xAI, Image model)](https://melius.com/grok-imagine): xAI's Grok image model for generation and editing
- [Veo 3.1 (Google, Video model)](https://melius.com/veo-3.1): Google's flagship Veo video model
- [Gemini Omni Flash (Google, Video model)](https://melius.com/gemini-omni-flash): Google's fast frontier video model
- [Seedance 2.0 (ByteDance, Video model)](https://melius.com/seedance-2.0): ByteDance's flagship video model with native audio
- [Seedance 2.0 Mini (ByteDance, Video model)](https://melius.com/seedance-2.0-mini): The fast, lightweight Seedance 2.0 video model
- [Sora 2 (OpenAI, Video model)](https://melius.com/sora-2): OpenAI's Sora 2 video model
- [Kling 3 Pro (Kuaishou, Video model)](https://melius.com/kling-3-pro): Kuaishou's high-quality Kling 3 video model
- [Kling 3 Turbo Pro (Kuaishou, Video model)](https://melius.com/kling-3-turbo): The faster Kling 3 Turbo Pro video model
- [Kling 3 Turbo Standard (Kuaishou, Video model)](https://melius.com/kling-3-turbo-standard): Kling 3 Turbo at standard quality for fast video
- [Wan 2.7 (Alibaba, Video model)](https://melius.com/wan-2.7): Alibaba's Wan 2.7 video model
- [Wan 2.2 (Alibaba, Video model)](https://melius.com/wan-2.2): Alibaba's fast Wan 2.2 image-to-video model
- [LTX 2.5 (Lightricks, Video model)](https://melius.com/ltx-2.5): Lightricks' latest LTX video model
- [LTX 2.3 (Lightricks, Video model)](https://melius.com/ltx-2.3): Lightricks' LTX video model with an audio-driven mode
- [LTX 2.3 Quality (Lightricks, Video model)](https://melius.com/ltx-2.3-quality): The high-quality LTX 2.3 image-to-video tier
- [Flux 3 (Black Forest Labs, Video model)](https://melius.com/flux-3): Black Forest Labs' frontier FLUX.3 video model
- [Flux 3 Extend (Black Forest Labs, Video model)](https://melius.com/flux-3-extend-video): Continue a clip with Black Forest Labs' Flux 3
- [Flux 3 Keyframes (Black Forest Labs, Video model)](https://melius.com/flux-3-keyframes): Animate between keyframes with Black Forest Labs' Flux 3
- [Grok Imagine Video (xAI, Video model)](https://melius.com/grok-imagine-video): xAI's fast Grok video model
- [Grok Imagine Video 1.5 (xAI, Video model)](https://melius.com/grok-imagine-video-1.5): xAI's Grok 1.5 image-to-video model
- [Kling 3.0 Omni (Kuaishou, Video model)](https://melius.com/kling-o3): Kuaishou's multimodal Kling 3.0 Omni video model
- [Kling 2.5 Turbo (Kuaishou, Video model)](https://melius.com/kling-2.5-turbo): The fast, popular Kling 2.5 Turbo video model
- [Kling 2.6 Pro (Kuaishou, Video model)](https://melius.com/kling-2.6-pro): Kuaishou's high-quality Kling 2.6 Pro video model
- [Happy Horse 1.1 (Alibaba, Video model)](https://melius.com/happy-horse-1.1): Alibaba's Happy Horse 1.1 video model
- [Vidu Q3 (Vidu, Video model)](https://melius.com/vidu-q3): Vidu's Q3 video model
- [Grok Imagine Video Extend (xAI, Video model)](https://melius.com/grok-imagine-video-extend): Extend an existing clip with xAI's Grok video model
- [Ray 3.2 (Luma, Video model)](https://melius.com/ray-3.2): Luma's Ray 3.2 video model
- [PixVerse v5.6 (PixVerse, Video model)](https://melius.com/pixverse-v5.6): PixVerse's v5.6 video model
- [Infinitalk (MultiTalk, Video model)](https://melius.com/infinitalk): Turn a photo and an audio clip into a talking avatar
- [HeyGen Lipsync (HeyGen, Video model)](https://melius.com/heygen-lipsync): HeyGen's lip-sync model for matching video to audio
- [Kling Lipsync (Kuaishou, Video model)](https://melius.com/kling-lipsync): Kuaishou's Kling lip-sync model
- [Sync Lipsync v3 (Sync, Video model)](https://melius.com/sync-lipsync-v3): Sync's latest lip-sync model
- [Sync Lipsync v2 Pro (Sync, Video model)](https://melius.com/sync-lipsync-v2-pro): Sync's high-quality v2 Pro lip-sync model
- [Kling Avatar (Kuaishou, Video model)](https://melius.com/kling-avatar): Kuaishou's Kling talking-avatar model
- [AI Avatar (MultiTalk, Video model)](https://melius.com/ai-avatar): Turn a portrait and audio into a talking avatar
- [VEED Fabric 1.0 (VEED, Video model)](https://melius.com/veed-fabric-1.0): VEED's Fabric model for talking-avatar video
- [VEED Lipsync v2 (VEED, Video model)](https://melius.com/veed-lipsync-v2): VEED's v2 lip-sync model
- [ElevenLabs Turbo v2.5 (ElevenLabs, Audio model)](https://melius.com/eleven_turbo_v2_5): ElevenLabs' fast, low-latency text-to-speech model
- [ElevenLabs Multilingual v2 (ElevenLabs, Audio model)](https://melius.com/eleven_multilingual_v2): ElevenLabs' multilingual text-to-speech model
- [ElevenLabs v3 (ElevenLabs, Audio model)](https://melius.com/eleven_v3): ElevenLabs' most expressive text-to-speech model
- [Seed Audio 1.0 (ByteDance, Audio model)](https://melius.com/seed-audio-1.0): ByteDance's Seed Audio model for reference-guided speech
- [ElevenLabs Sound Effects (ElevenLabs, Audio model)](https://melius.com/elevenlabs-sfx): Generate sound effects from a text prompt
- [ElevenLabs Music (ElevenLabs, Audio model)](https://melius.com/elevenlabs-music): Compose original music from a text prompt
- [MiniMax Music 3 (MiniMax, Audio model)](https://melius.com/minimax-music-3): Generate complete songs with vocals from a description
- [ElevenLabs Voice Changer (ElevenLabs, Audio model)](https://melius.com/elevenlabs-voice-changer): Convert one voice into another
- [ElevenLabs Voice Isolator (ElevenLabs, Audio model)](https://melius.com/elevenlabs-voice-isolator): Isolate a clean voice from noisy audio
- [GPT-5.6 (OpenAI, Text model)](https://melius.com/gpt-5.6): OpenAI's flagship GPT-5.6 text model
- [Grok 4.6 (xAI, Text model)](https://melius.com/grok-4.6): xAI's flagship Grok 4.6 text model
- [Grok 4.5 (xAI, Text model)](https://melius.com/grok-4.5): xAI's Grok 4.5 text model
- [DeepSeek V4 Pro (DeepSeek, Text model)](https://melius.com/deepseek-v4-pro): DeepSeek's V4 Pro reasoning text model
- [Llama 4 Maverick (Meta, Text model)](https://melius.com/llama-4-maverick): Meta's open Llama 4 Maverick text model
- [Claude Haiku 4.5 (Anthropic, Text model)](https://melius.com/claude-haiku-4.5): Anthropic's fast Claude Haiku 4.5 model
- [Mistral Large 3 (Mistral, Text model)](https://melius.com/mistral-large-3): Mistral's flagship Large 3 text model
- [Qwen 3.6 Plus (Alibaba, Text model)](https://melius.com/qwen3.6-plus): Alibaba's multimodal Qwen 3.6 Plus text model
- [Gemini 3.1 Pro (Google, Text model)](https://melius.com/gemini-3.1-pro): Google's flagship Gemini 3.1 Pro model
- [Gemini 3.5 Flash (Google, Text model)](https://melius.com/gemini-3.5-flash): Google's fast multimodal Gemini 3.5 Flash model
- [Claude Sonnet 5 (Anthropic, Text model)](https://melius.com/claude-sonnet-5): Anthropic's balanced Claude Sonnet 5 model
- [Claude Opus 4.8 (Anthropic, Text model)](https://melius.com/claude-opus-4.8): Anthropic's most capable Claude Opus 4.8 model
- [Qwen 3 VL 235B (Alibaba, Text model)](https://melius.com/qwen3-vl-235b): Alibaba's Qwen 3 VL vision-language model
- [Qwen 3.5 397B (Alibaba, Text model)](https://melius.com/qwen3.5-397b): Alibaba's largest Qwen 3.5 multimodal model

## Guides and articles
- [Seedance 2.5 is live on Melius: 30-second video with audio](https://melius.com/blog/seedance-2-5): Seedance 2.5 is live on Melius. ByteDance's new video model generates up to 30 seconds in a single take with audio in the same pass, and takes up to 50 reference files. Here is what shipped, what you can run today, and what is still coming.
- [The best Figma Weave alternative for generative design](https://melius.com/blog/figma-weave-alternative): Looking for a Figma Weave (Weavy) alternative? Weave puts a generation canvas next to your design files, great if you live in Figma. If you don't, the ecosystem coupling and fast credit burn become constraints. Here's how Melius compares as a standalone, canvas-first workspace with an agent and one credit pool.
- [The best Magnific alternative for upscaling and creative AI](https://melius.com/blog/magnific-alternative): Looking for a Magnific alternative? Magnific's creative upscaler is a strong specialist, but if upscaling is one step in a real production, not the whole tool, and you'd rather not pay premium rates for a no-refund, credit-metered enhancer, here's how Melius folds generation, editing, resizing and assembly onto one canvas.
- [The best Flora alternative for creative teams](https://melius.com/blog/flora-alternative): Looking for a Flora alternative? Flora (Florafauna) and Melius are the two node canvases most often compared, so "we have a canvas and an agent" isn't the difference. The real gaps are collaboration, learning curve, and pricing. Here's how Melius handles the same work with true real-time multiplayer.
- [The best Higgsfield alternative for cinematic AI video](https://melius.com/blog/higgsfield-alternative): Looking for a Higgsfield alternative? Higgsfield nails cinematic camera motion, but teams outgrow its fixed roster and export-to-edit workflow. Here's an honest look at how Melius handles the same work, with every frontier model, an agent that plans multi-step pipelines, and a place to finish the cut.
- [The modern creative workflow: every tool you need, in one place](https://melius.com/blog/modern-creative-workflow-tools): The modern creative stack has sprawled into a dozen subscriptions, stock sites, and disconnected apps. Here is a guide to the tools, assets, and platforms behind a fast creative workflow, and how a single canvas like Melius collapses generation, editing, assets, and collaboration into one place.
- [AIGC for marketing: scaling AI-generated visual content](https://melius.com/blog/aigc-marketing-content): What is AIGC? AI-generated content is reshaping how marketing teams produce images and video at scale. Here is a plain-English guide to generative AI in content workflows, how Google views AI content, and how a single canvas like Melius turns one brief into on-brand visuals across every model.
- [Connect Claude to a creative canvas with MCP](https://melius.com/blog/connect-ai-agents-canvas-mcp): Connect Claude to a creative canvas over MCP and skip the context drift you get from copy-pasting into a chatbot. The Model Context Protocol exposes the Melius canvas as live, structured data an agent can read and act on. Here is how to connect Claude to Melius, what the tools actually are, and why the loop self-corrects.
- [One subscription, every model: an all-in-one AI aggregator guide](https://melius.com/blog/one-subscription-every-model): An AI aggregator puts every model, text, image, video and audio, under one subscription instead of a fragmented stack that clears $100 a month. Here is why aggregators are winning, what a true all-in-one creative AI platform must deliver, and a three-step plan to consolidate your workflow onto a single canvas.
- [The Brief-to-Execution workflow for creative teams](https://melius.com/blog/brief-to-execution-workflow): One-shot prompting has a ceiling. The agencies that win aren't the ones with the best prompt. They're the ones that turn a single master brief into a full, on-brand campaign. Here is the Brief-to-Execution workflow, and how an orchestration layer like Melius splits the Labor of Execution from the Labor of Taste.
- [Seedance 2.5 is coming: ByteDance's next-gen AI video model](https://melius.com/blog/seedance-2-5-anticipation): Seedance is behind more than a quarter of every video generation on Melius, and its share keeps climbing. ByteDance's Seedance 2.5 is announced with 50 reference inputs, synced audio, and, per its makers, native 4K. Here is the real usage data behind Seedance today, and an honest read on what 2.5 does and does not confirm.
- [Which Nano Banana model should you use?](https://melius.com/blog/nano-banana-image-models): Google's Nano Banana is three image models, not one. We pulled real generation data from Melius to work out when to use Nano Banana Pro, Nano Banana 2, and Nano Banana 2 Lite.

## Resources
- [Web app](https://app.melius.com): create on the Melius canvas; free to start and you only pay for what you generate
- [Documentation](https://docs.melius.com)
- [RSS feed](https://melius.com/blog/feed.xml)
- [Privacy policy](https://melius.com/privacy)
- [Terms of service](https://melius.com/terms)