Add one URL to Claude, ChatGPT or Cursor and ask for what you need. 496 models from Google, OpenAI, Black Forest Labs, Kling, ByteDance and ElevenLabs sit behind it, and the router picks the right one for every job. You pay per result.
https://api.market/api/mcp/gatewayFree to add. No subscription. Everything on this page was made through it, shown with its real price.
Models fromGoogleOpenAIBlack Forest LabsKlingByteDanceAlibabaxAIElevenLabsMiniMaxIdeogramRecraft
https://api.market/api/mcp/gatewayIt is the API.market MCP server. Free to add, and you sign in once. No API key to paste.
“Make a hero image for my blog post about remote work, 16:9.”
$0.042Your agent picks the model, runs the job and hands back the file. You pay per result from your API.market wallet.
Who it is for
You do not need to know which model is best this month. Add API.market to the agent you already use, or call one API, and ask for the result you want.
Ask your agent for a style, a logo or a pattern and it picks the model that is best at that kind of work. Oil paint, anime, 3D, paper craft or clean vector, without learning which model does what.
Just ask
Images
Each one is a single ask to an agent, or one API call. They cost from $0.013 each. Hover for the prompt. Click to see it full size, with the model that made it and the words to reuse.
Edit what you have
Stage an empty room, restore an old print, change the season or the color. Drag the line to compare what went in with what came back.
Virtually stage this exact empty room for a real estate listing: add a light grey sofa, a round oak coffee table, a wool rug, a floor lamp, plants and framed art. Keep the walls, floor, window and lighting unchanged.
Video
Text to video, photo to video and talking presenters. Every clip is a real job and is playing now. Pick one to watch it large. Clips with a voice or a soundtrack have a sound button.
A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
Beyond pixels
The same URL does more than pictures. Press play on the voiceover, read what came back from a receipt, spin a 3D model. Every card shows what was asked and what it cost.
POST /v1/3d
{
"model": "firtoz/trellis",
"input_references": [
{
"type": "image_url",
"role": "image",
"image_url": {
"url": "https://api.market/media-api/background-removal.webp"
}
}
],
"parameters": {
"generate_model": true
}
}A textured GLB mesh from a single cutout. Drag to turn it.
{ "store": "Blue Door Cafe", "date": "2026-08-11", "time": "09:42", "items": [ { "name": "Flat white", "price": 4.20 }, { "name": "Oat croissant", "price": 3.80 }, { "name": "Sparkling water", "price": 2.50 } ], "subtotal": 10.50, "tax": 0.84, "total": 11.34 }Prompt: Read this receipt. Return only JSON: {"store":..., "date":..., "time":..., "items":[{"name":...,"price":...}], "subtotal":..., "tax":..., "total":...}
POST /v1/understand
{
"model": "perceptron-ai-inc/isaac-0.1",
"prompt": "Read this receipt. Return only JSON: {\"store\":..., \"date\":..., \"time\":..., \"items\":[{\"name\":...,\"price\":...}], \"subtotal\":..., \"tax\":..., \"total\":...}",
"input_references": [
{
"type": "image_url",
"role": "image",
"image_url": {
"url": "https://api.market/media-api/receipt.webp"
}
}
]
}Ask for the fields you want and get structured JSON back.
Meet the new trail runner, light enough for race day, tough enough for every day. Order today and get free returns for 30 days.
Prompt: Meet the new Trail Runner. Light enough for race day, tough enough for every day. Order today and get free returns for thirty days.
POST /v1/audio
{
"model": "elevenlabs/v3",
"prompt": "Meet the new Trail Runner. Light enough for race day, tough enough for every day. Order today and get free returns for thirty days."
}ElevenLabs reads it; speech-to-text returns every word with a timestamp.
POST /v1/3d
{
"model": "firtoz/trellis",
"input_references": [
{
"type": "image_url",
"role": "image",
"image_url": {
"url": "https://api.market/media-api/mug-cutout.webp"
}
}
],
"parameters": {
"generate_model": true
}
}Any product photo becomes a 3D asset for AR or games.
Alt text: White and green running shoe with mesh upper and white sole. Product description: A stylish white and green running shoe featuring a breathable mesh upper and a white sole, perfect for both comfort and performance.
Prompt: Write SEO alt text (under 125 characters) and a one-sentence product description for this shoe photo.
POST /v1/understand
{
"model": "perceptron-ai-inc/isaac-0.1",
"prompt": "Write SEO alt text (under 125 characters) and a one-sentence product description for this shoe photo.",
"input_references": [
{
"type": "image_url",
"role": "image",
"image_url": {
"url": "https://api.market/media-api/product-photo.webp"
}
}
],
"models": [
"yorickvp/llava-13b",
"csviverdeia/locateanything-3b-h100"
]
}Accessible alt text and a product line from one photo.
Prompt: Warm upbeat lo-fi jingle for a neighbourhood coffee brand, mellow electric piano, soft brushed drums, gentle bass, cozy morning mood
POST /v1/audio
{
"model": "google/lyria-2",
"prompt": "Warm upbeat lo-fi jingle for a neighbourhood coffee brand, mellow electric piano, soft brushed drums, gentle bass, cozy morning mood"
}A minute of original music from one sentence.
Why you need this
Every model has its own API, its own price and its own strengths. Pick the wrong one and you pay far more, wait far longer, or get a worse result. Nobody has time to keep up, so the router does it for you.
We sent one prompt to 12 image models
Same request. 242x difference in price, 29x in speed.
One more model failed at the provider and cost nothing. With the router a backup takes over.
Doing it yourself
With API.market
Smart routing
You describe the job in plain words. Jev, our judge model, reads it, the router sets aside every model that is not good enough, ranks the rest by what matters to you, and lines up backups. Pick a job and a priority and watch it choose.
5/5Best fit wins. Jev scores how well each model fits your job. The router ranks them by your priority using price, measured speed and success rate, runs the winner and keeps up to two backups ready.
Click any model to see what it made and why it was or was not picked.
Ranked by balancedPickedBackups
POST /v1/images
{
"model": "auto",
"use_case": "photo for a blog post header",
"priority": "balanced",
"prompt": "A tiny glass greenhouse café on a rainy Tokyo si..."
}Backups: black-forest-labs/flux-3-image then xai/grok-imagine-image-2. If the pick cannot start, the next one runs and the failed attempt costs nothing. Click any model on the left to see what it made.
Quality tiers come from public leaderboards (Artificial Analysis, LMArena). A weak model, or a community fine-tune with no rating, never wins on price or speed alone. On live requests the router also checks how well each model fits your use case.
What makes it smart
You say what you want. The router works out what kind of job it is, which models are good at it, and which one fits your budget or your deadline. These are its real answers, asked on 10 November 2026.
Pick a request and see what the router decided.
Backups: men1scus/birefnet, 851-labs/background-remover
“A product photo for an online store listing”, asked four ways. Three different models win.
The highest rated model that can do the job.
Top tier and mid-priced, so it wins on balance too.
The lowest price among models that clear the quality bar.
The quickest measured run time on our network.
Ask for a 5 second clip and the length is set for you. You do not need to learn each model's options.
Most picks come with up to two backups of the same kind. If the first model cannot start, the next one runs without you doing anything.
If a model refuses your job or errors out before it runs, that attempt costs nothing. Every job shows what it cost.
Set the most a job may cost. Models that usually cost more are left out, and the hold on your wallet stops at the cap. If a pay as you go run still goes over, the difference comes back, rounded down to the cent.
Run times and success rates come from real jobs on our network over the last 30 days.
Weak, outdated and unrated community models start at the back and need a strong fit score to be picked, however cheap they are.
How a job runs
Real jobs, step by step. Your wallet is checked first, backups wait in line, and a model that cannot start costs nothing. Pick a job to replay it.
A white and sage-green running sneaker, three-quarter side view, on a seamless pure white background, soft studio shadow, sharp product photography
Add it to your agent
Pick your app and follow the steps. It takes about a minute. You sign in to API.market once, so there is no API key to paste. Your agent gets media_find_models, media_generate and media_get_job, plus every other API on the marketplace, billed to your wallet.
https://api.market/api/mcp/gatewayFree Claude plans allow one custom connector. On Team and Enterprise an owner adds it once under Organization settings, then members connect.
Another app? Setup for every MCP client
Real requests and the results they produced. The agent picks the model, runs the job and hands back the file.
$0.042“Make a hero image for my blog post about remote work, 16:9.”
$0.03“Shoot our sage-green sneaker as a clean product photo on white.”
“Turn this street photo into a 5 second video with gentle motion.”
“Voice a 10 second ad for the Trail Runner, warm and upbeat.”
$0.0022“Remove the background from this product shot.”
{ "store": "Blue Door Cafe", "date": "2026-08-11", "time": "09:42", "items": [ { "name": "Flat white", "price": 4.20 }, { "name": "Oat croissant", "price": 3.80 }, { "name": "Sparkling water", "price": 2.50 } ], "subtotal": 10.50, "tax": 0.84, "total": 11.34 }$0.0028“Read this receipt and give me the items and total as JSON.”
“Design a minimal coffee brand logo I can edit, as SVG.”
“Make this portrait read our launch script as a talking video.”
What the router can do
Images, video, audio, understanding and 3D. Tell your agent what you want, or call the endpoint yourself. Click any job to see the models that do it and what they cost.
/v1/imagesMake new images, edit photos, remove backgrounds, upscale, restore, swap faces and try clothes on.
“Make a skincare ad with the headline GLOW DAILY”
/v1/videosTurn a line of text or a photo into video, make a portrait talk, lip-sync, restyle and upscale clips.
“Make a 5 second cinematic latte pour”
/v1/audioVoiceovers, cloned voices, original music and stem separation.
“Voice this 10 second ad, warm and upbeat”
/v1/understandTranscribe audio, read receipts and documents, caption and detect what is in a picture.
“Read this receipt and give me the items as JSON”
/v1/3dTurn a product photo or a sentence into a textured 3D model.
“Turn this product photo into a 3D model”
Every task
Pick an endpoint and open any job to see the well-rated models that do it, what they cost, the words to give your agent and the request to send.
POST /v1/images
{
"model": "auto",
"use_case": "Make a hero image for my blog post about remote work",
"prompt": "Make a hero image for my blog post about remote work",
"priority": "balanced"
}Model shootout
We sent the same prompt to the leading image and video models. Sort by leaderboard rank, price or run time and judge the results yourself. Every price and time is what that job really cost and took.
Prompt: A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo
Prompt: A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo
POST /v1/images
{
"model": "openai/gpt-image-2.5-sunburst",
"prompt": "A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo",
"aspect_ratio": "1:1"
}Prompt: A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo
POST /v1/images
{
"model": "openai/gpt-image-2",
"prompt": "A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo",
"aspect_ratio": "1:1"
}Prompt: A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo
POST /v1/images
{
"model": "google/nano-banana-2.1",
"prompt": "A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo",
"aspect_ratio": "1:1"
}Prompt: A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo
POST /v1/images
{
"model": "xai/grok-imagine-image-2",
"prompt": "A tiny glass greenhouse café on a rainy Tokyo side street at dusk, warm window light, reflections on wet pavement, a cyclist with an umbrella passing, 35mm film photo",
"aspect_ratio": "1:1"
}Prompt: A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
Prompt: A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
POST /v1/videos
{
"model": "alibaba/wan-3",
"prompt": "A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic",
"aspect_ratio": "16:9",
"parameters": {
"duration": 5
}
}Prompt: A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
POST /v1/videos
{
"model": "bytedance/seedance-2.0",
"prompt": "A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic",
"aspect_ratio": "16:9",
"resolution": "720p",
"parameters": {
"duration": 5,
"generate_audio": false
}
}Prompt: A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
POST /v1/videos
{
"model": "prunaai/p-video-2-pro",
"prompt": "A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic",
"aspect_ratio": "16:9"
}Prompt: A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic
POST /v1/videos
{
"model": "kwaivgi/kling-v3-video",
"prompt": "A barista pours steamed milk into a latte, forming a leaf pattern, slow motion close-up, warm morning window light, shallow depth of field, cinematic",
"aspect_ratio": "16:9"
}Popular models
Public leaderboard rank, our price for a typical job, and the run time and success rate we measured over the last 30 days.
For developers
Your code calls the Media API. Your agent calls the MCP server, which runs on the same router. Same models, same prices, same wallet.
https://router.api.marketPOST to /v1/images, /v1/videos, /v1/audio, /v1/understand or /v1/3d with Authorization: Bearer and your API.market key. Name a model, or send model auto and let the router pick. Each endpoint lists its models (/v1/images/models and so on), and /playground lets you try requests in the browser.
https://api.market/api/mcp/gatewayAdd it to Claude, ChatGPT, Cursor, VS Code, Claude Code or Codex and sign in with OAuth. Your agent gets media_find_models, media_generate and media_get_job, which run on the same router, plus every other API on the marketplace.
Full reference: OpenAPI, llms.txt and the playground.
curl https://router.api.market/v1/images \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "auto",
"use_case": "e-commerce product photo on white",
"prompt": "A white and sage-green running sneaker on pure white",
"priority": "balanced",
"aspect_ratio": "1:1"
}'Get an API key
Make an account, add a few dollars, and paste one command. Your first product photo costs about $0.03, so $5 covers more than a hundred.
Sign in with Google, GitHub or email. No subscription.
Create accountPay as you go. The same wallet pays for every API on API.market.
Paste it in a terminal after you add your key.
curl https://router.api.market/v1/images -H 'Authorization: Bearer YOUR_API_KEY' -H 'Content-Type: application/json' -H 'Prefer: wait=45' -d '{"model":"black-forest-labs/flux-3-image","prompt":"A white and sage-green running sneaker, three-quarter side view, on a seamless pure white background, soft studio shadow, sharp product photography","aspect_ratio":"1:1","models":["google/nano-banana-2.1","alibaba/qwen-image-3"]}'One API for image generation, image editing, video, voice, music, transcription, OCR and 3D. You call router.api.market with your API.market key and either name a model or set model to auto and describe the job.
Each model is billed at its pay-as-you-go price on API.market: per image, per second of video, per 100 characters of speech, or per run. A product photo on FLUX 3 costs about $0.03. There is no subscription; you top up your wallet from $5.
Create a free API.market account, add credit to your wallet, copy your API key from the account page, and send a POST request to https://router.api.market/v1/images with the header Authorization: Bearer YOUR_API_KEY.
Image models from Google (Nano Banana), OpenAI (GPT Image), Black Forest Labs (FLUX), ByteDance (Seedream), Ideogram, Recraft, Alibaba and xAI; video from Kling, Seedance, Wan, MiniMax Hailuo, PixVerse and LTX; audio from ElevenLabs, MiniMax, Google Lyria and xAI; plus open models for transcription, OCR, upscaling, try-on and 3D.
No. Name a model and the job runs on that model. You can also list fallback models in your own order, or set model to auto with a use case and a priority of cost, speed or quality.
If a model cannot start your job, the next model on your list takes it and you pay only for the model that ran. When a model fails, the job returns the error and what, if anything, it cost.
Yes. Add the MCP server at https://api.market/api/mcp/gateway to Claude, ChatGPT, Cursor, VS Code, Claude Code or Codex and sign in with OAuth. Your agent gets media_find_models, media_generate and media_get_job, billed to the same wallet.
router.api.market is the Media API itself: your code sends POST /v1/images, /v1/videos, /v1/audio, /v1/understand or /v1/3d there with your API key. https://api.market/api/mcp/gateway is the MCP server for AI agents; its media tools call router.api.market for you with your key. Same models, same prices, same wallet.
No. Add the MCP server to Claude, ChatGPT, Cursor or VS Code, sign in once, and ask in plain words: make a product photo, turn this picture into a video, voice this script. Designers, advertisers and video teams use it this way. Developers can call the same models over the API.
Jev is our judge model. When you send model auto with a use case, Jev works out the task and settings and scores how well each model fits the job. Models rated weak or outdated on public leaderboards, and community fine-tunes with no rating, start at the back. The router ranks the models by your priority (quality, price, speed or balanced) using price, measured run time and success rate. It runs the winner and keeps backups ready.
Output links work for 24 hours. Download anything you want to keep.
Copy the URL, add it as a custom MCP server and ask for what you need. Images start at $0.0016, and a model that cannot start your job costs nothing.
https://api.market/api/mcp/gatewayBuilding it into your own product? Get an API key