All Image LabsImage
Text to image across the full OpenRouter image roster. Control aspect ratio, resolution up to 4K, seed, and negatives. Or edit and reframe an existing shot.
MCP server · bring your own key
Generate image, video & music from your AI assistant.
All Image Labs plugs into Claude, Cursor, and any MCP client. Ask in plain language and it runs 50+ generative models on your own OpenRouter key. No account, nothing stored.
uvx studio-mcp
Stateless & private7 tools, 3 modalitiesRuns on your machine
modalities image · video · musictools 7models 50+data stored 0install uvx studio-mcp
Three modalities, one connection.
Every tool maps to a plain-language request. Your assistant picks the model, fills the parameters, and hands back the result: image inline, video and audio saved to disk.
Video
Text and image to video on Veo, Kling, Seedance, Sora and more. Drive motion with a start frame, interpolate to an end frame, or carry a style reference. Native audio where the model supports it.
Music
Score with Google Lyria 3: a quick clip for a scene or a full track, shaped by mood and genre. Files land in your output folder, ready to drop into an edit.
enhance_prompt
Turn a terse idea into a rich, model-ready prompt before you spend a credit.
list_models
See every generation model your OpenRouter key can reach, grouped by modality.
get_capabilities
Resolutions, durations, aspect ratios, audio and frame support for any model.
estimate_cost
Get the credit cost of a generation before your assistant runs it.
Connected in about a minute.
Paste one block into your MCP client's config, restart it, and ask. No build step, no SDK. uvx fetches and runs the server on demand.
Get an OpenRouter key
Create a key at openrouter.ai/keys. This is what pays for and runs every model.
Add the server to your client
Drop the config on the right into Claude Desktop, Cursor, or any MCP client, then restart it.
Ask your assistant
“Generate a 16:9 cinematic shot of a lighthouse at dusk, then animate it with Veo.” That's it.
{"mcpServers": {"all-image-labs": {"command": "uvx","args": ["studio-mcp"],"env": {"OPENROUTER_API_KEY": "sk-or-v1-…"}}}}
Run the models you already read about.
Whatever OpenRouter ships, your assistant can call. A few of what's reachable today:
Video
Veo 3.1 google/veo-3.1Kling 2.5 kwaivgi/klingSeedance 2.0 bytedance/seedanceSora 2 Pro openai/sora-2-proHailuo 2.3 minimax/hailuoWAN 2.5 alibaba/wanGrok Imagine x-ai/grok-imagineVeo 3.1 google/veo-3.1Kling 2.5 kwaivgi/klingSeedance 2.0 bytedance/seedanceSora 2 Pro openai/sora-2-proHailuo 2.3 minimax/hailuoWAN 2.5 alibaba/wanGrok Imagine x-ai/grok-imagineVeo 3.1 google/veo-3.1Kling 2.5 kwaivgi/klingSeedance 2.0 bytedance/seedanceSora 2 Pro openai/sora-2-proHailuo 2.3 minimax/hailuoWAN 2.5 alibaba/wanGrok Imagine x-ai/grok-imagineVeo 3.1 google/veo-3.1Kling 2.5 kwaivgi/klingSeedance 2.0 bytedance/seedanceSora 2 Pro openai/sora-2-proHailuo 2.3 minimax/hailuoWAN 2.5 alibaba/wanGrok Imagine x-ai/grok-imagine
Image
Gemini 3 Pro Image google/gemini-3-pro-imageNano Banana google/gemini-2.5-flash-imageFLUX.2 black-forest-labs/flux.2GPT-Image openai/gpt-image-1Seedream 4 bytedance/seedreamRecraft v3 recraft/recraft-v3Riverflow sourceful/riverflowGemini 3 Pro Image google/gemini-3-pro-imageNano Banana google/gemini-2.5-flash-imageFLUX.2 black-forest-labs/flux.2GPT-Image openai/gpt-image-1Seedream 4 bytedance/seedreamRecraft v3 recraft/recraft-v3Riverflow sourceful/riverflowGemini 3 Pro Image google/gemini-3-pro-imageNano Banana google/gemini-2.5-flash-imageFLUX.2 black-forest-labs/flux.2GPT-Image openai/gpt-image-1Seedream 4 bytedance/seedreamRecraft v3 recraft/recraft-v3Riverflow sourceful/riverflowGemini 3 Pro Image google/gemini-3-pro-imageNano Banana google/gemini-2.5-flash-imageFLUX.2 black-forest-labs/flux.2GPT-Image openai/gpt-image-1Seedream 4 bytedance/seedreamRecraft v3 recraft/recraft-v3Riverflow sourceful/riverflow
Music
Lyria 3 Pro google/lyria-3-proLyria 3 Clip google/lyria-3-clipLyria 3 Pro google/lyria-3-proLyria 3 Clip google/lyria-3-clipLyria 3 Pro google/lyria-3-proLyria 3 Clip google/lyria-3-clipLyria 3 Pro google/lyria-3-proLyria 3 Clip google/lyria-3-clip
Private by design
Your keys and your prompts stay yours.
- Your OpenRouter key is used per request and never stored or logged on our side.
- No account, no database, no cloud copies. Generated files land on your machine.
- Optional usage counting is anonymous (model and success only) and switches off with one env var.
Questions, answered.
All Image Labs is an MCP server that adds image, video, and music generation to any AI assistant that speaks the Model Context Protocol, including Claude Desktop and Cursor. You connect it once with your own OpenRouter key, then ask your assistant to generate media in plain language. It runs on your machine; there is no separate app to learn.
The Model Context Protocol is an open standard for giving AI assistants new tools. Any MCP client works: Claude Desktop, Cursor, Cline, Zed, and the growing list of MCP-capable apps. Adding All Image Labs gives your assistant seven new tools for generating and shaping media.
No. There is no sign-up, no dashboard, and no password. The server is stateless: it holds no accounts and no database. You bring an OpenRouter key, the server uses it for the length of a single request, and that is the entire relationship.
Your OpenRouter key is read from your local MCP client config and passed straight to OpenRouter per request. It is never written to disk on our side and never logged. Prompts and generated media go between your machine and OpenRouter only. We never see your content.
Any generation model on OpenRouter, across all three modalities: video (Veo 3.1, Kling, Seedance, Sora 2, Hailuo, WAN, Grok), image (Gemini 3 Pro Image, Nano Banana, FLUX.2, GPT-Image, Seedream, Recraft, Riverflow), and music (Google Lyria 3). New models appear in your assistant automatically as OpenRouter adds them.
The MCP server itself is free and open to use. You pay OpenRouter directly for the model calls you make, at their published per-generation prices. Call
estimate_cost first and your assistant will tell you the price before it runs anything.Nothing. Images come back inline; video and audio are written to a folder on your own machine. The only thing that ever leaves is an optional anonymous counter (model name and a success flag, never prompts, content, or keys), and you can switch it off with
STUDIO_MCP_TELEMETRY=0.Give your assistant a camera and a studio.
Add All Image Labs to your MCP client and generate image, video, and music without leaving the chat.