Make videos with any AI bot

Three ways in

1. Video by link — works with every bot. The bot writes a short JSON recipe into https://indiadigital.org/video/make/#…; opening it renders the video, then showsDownload and Share.
2. Remote MCP at https://indiadigital.org/mcp — for bots that can add an MCP server (Grok, Claude, ChatGPT connectors, Muse Code). Recipes can include Pixabay stock and AI-generation steps. Tools: video_guide,list_video_templates, list_video_styles, get_edit_plan_schema, validate_edit_plan, build_video_link. No login, no media, read-only; it returns links.
3. Local bridge (npx indiadigital-video-mcp) — full control of the user’s own editor tab for agents on the user’s computer: import footage, analyse shots, edit, review, export. Set it up.

Set up your bot

Grok (xAI)

1. In Grok, open the place where you add tools or connectors and add a remote MCP server with the URL https://indiadigital.org/mcp (no authentication).
2. Ask: “Make a 30-second Diwali reel for my shop, Sharma Sweets, offer 20% off, phone 9876543210. Use IndiaDigital: build the video link and give it to me.”
3. Grok calls build_video_link and replies with a link. Open it; the video renders in your browser.

Note: Grok runs in the cloud, so it cannot reach the local bridge on your computer (that would need a tunnel). Use the remote MCP or video by link.

Meta AI (Muse)

1. Give Muse the MCP URL and ask it to make a skill. Paste this:

   Make a skill called "IndiaDigital video". When I ask for a video (festival greeting, shop offer, birthday, invite, reel): 1) call the MCP server https://indiadigital.org/mcp → video_guide once, 2) pick a template with list_video_templates and fill its fields from my request, 3) call build_video_link and give me the link, 4) if you can open links, open the agentLink, wait until the page shows "Your video is ready", and download it.

2. Or skip the setup and use video by link — paste this once:

   To make a video, write a JSON recipe and give me this link: https://indiadigital.org/video/make/#q=<URL-encoded JSON>. Recipe: {"template":"<id>","fields":{…},"media":["https://…"],"aspect":"9:16"} or just {"brief":"<what I asked>"}. Templates and fields: https://indiadigital.org/video/for-bots/llms.txt

3. Muse Code can add https://indiadigital.org/mcp as an MCP server directly.

ChatGPT and OpenAI Dots

1. Video by link works today: paste the instruction above, then ask for your video. ChatGPT writes the link; open it.
2. If your ChatGPT plan lets you add a custom MCP connector, add https://indiadigital.org/mcp; then it can call build_video_link itself.
3. Dots and other agents that browse can open the ?agent=1 link, wait for the ready state and download the file (see automation).

Claude, Claude Code, Cursor, Codex, Gemini CLI

• claude.ai and Claude Desktop: Settings → Connectors → Add custom connector → https://indiadigital.org/mcp.
• Claude Code (remote tools): claude mcp add --transport http indiadigital https://indiadigital.org/mcp
• Full control of your own footage (local bridge): claude mcp add indiadigital-video -- npx -y indiadigital-video-mcp, then open the editor and pair. Same idea for Cursor, Codex and Gemini CLI — configs on the agents page.

Copy-paste prompts

• Make a 30-second Diwali reel for my shop, Sharma Sweets, offer 20% off, phone 9876543210. Use IndiaDigital: build the video link and give it to me.
• Make a birthday video for Meera turning 7, from Papa, with these photos: <links>. Use IndiaDigital video by link.
• Turn these before/after clips into a 25-second Instagram reel for Aether Spaces with house music: <links>. Use the IndiaDigital before-after style.
• Make a Holi offer video for Gupta Garments: flat 30% off till 15 March, vertical 9:16.

The recipe

A recipe is a small JSON object. Give one of template (fill in the blanks), style (edit your clips), plan (a fulledit plan) — or just a brief in words and the page picks a template and fills what it can. Schema:/video/recipe-schema.json.

Link formats

• /video/make/#q= + encodeURIComponent(JSON) — easiest to write by hand.
• /video/make/#j= + base64url(JSON).
• /video/make/#r= + deflate-raw + base64url — shortest; what build_video_link returns.

Examples

Diwali offer for a shop (no photos needed)

{
  "template": "diwali",
  "fields": {
    "mode": "business",
    "shop": "Sharma Sweets",
    "offer": "20",
    "phone": "9876543210"
  },
  "aspect": "9:16"
}

Open this example

Just a brief

{
  "brief": "Make a 30-second Diwali reel for my shop, Sharma Sweets, offer 20% off"
}

Open this example

Birthday with your photos

{
  "template": "birthday-wish",
  "fields": {
    "name": "Meera",
    "age": "7",
    "from": "Papa"
  },
  "media": [
    "https://raw.githubusercontent.com/you/photos/main/meera1.jpg",
    "https://raw.githubusercontent.com/you/photos/main/meera2.jpg"
  ]
}

Open this example

Before/after reel from your clips

{
  "style": "before-after",
  "brief": "Aether Spaces interior transformations",
  "media": [
    {
      "url": "https://res.cloudinary.com/demo/video/upload/before1.mp4",
      "role": "before"
    },
    {
      "url": "https://res.cloudinary.com/demo/video/upload/after1.mp4",
      "role": "after"
    }
  ],
  "music": "house"
}

Open this example

Pixabay B-roll plus an AI intro shot (uses the viewer’s keys)

{
  "style": "product-launch",
  "brief": "Monsoon sale for Sharma Sweets: 20% off all mithai, vertical reel",
  "steps": [
    {
      "type": "generate",
      "mode": "t2v",
      "prompt": "slow push-in on a glowing sweet shop counter at dusk, rain outside, warm light, cinematic",
      "seconds": 5,
      "aspect": "9:16",
      "placement": "append",
      "count": 1
    },
    {
      "type": "stock",
      "query": "indian sweets",
      "kind": "video",
      "orientation": "vertical",
      "placement": "append",
      "count": 3
    },
    {
      "type": "stock",
      "query": "monsoon rain window",
      "kind": "video",
      "placement": "append",
      "count": 1
    }
  ],
  "music": "lofi",
  "aspect": "9:16"
}

Open this example

Festival greeting with stock photos

{
  "template": "diwali",
  "fields": {
    "mode": "personal",
    "from": "The Iyer family"
  },
  "steps": [
    {
      "type": "stock",
      "query": "diya lamps night",
      "kind": "photo",
      "count": 3,
      "placement": "append"
    }
  ]
}

Open this example

Templates and fields

For bots that click through pages (?agent=1)

• Open /video/make/?agent=1#…. Rendering starts on its own.
• Wait until <html data-video-state="ready"> (other states: idle, checking, working, error). #video-status is an aria-live status with data-progress (0–1).
• Click #download (it has data-ready="true", data-filename, data-bytes). #video-result holds the result as JSON.
• No link? Paste a recipe or plan into the labelled box #recipe-input and press #make-from-paste.
• Scripts: await window.idVideoMake.make(recipe), then window.idVideoMake.result.

Where the photos and clips come from

1. The bot brings its own media links. Put https links in media. The host must allow cross-site reads (see below).
2. Free stock from Pixabay. Add {"type":"stock","query":"…","kind":"photo|illustration|video","count":3} to steps. It uses the viewer’s own free Pixabay key: the page asks once and keeps it in that browser. Credits are kept with the video.
3. Creative Commons music, sound effects and images from Openverse (in the editor, no key): agents connected through thelocal bridge or WebMCP call search_openverse (kind music, sfx or image) andadd_openverse_to_timeline. Only CC0, public domain, CC BY and CC BY-SA, never NonCommercial or NoDerivatives. Title, author, source and licence are saved on the file. Call get_stock_credits and put the CC BY credits in the post caption. Add music before apply_style_template and the director cuts to that track instead of composing one. Share-alike (BY-SA) works are left out unless you pass noShareAlike: false.
4. AI generation. Add {"type":"generate","mode":"t2i|t2v|i2v","prompt":"…","seconds":5}. It runs with the viewer’s own setup: ComfyUI on their PC (free), “Use my ChatGPT subscription” (their official Codex CLI run by the local relay with --codex), or their own keys for fal.ai, kie.ai, Together or Google Gemini (Replicate and Higgsfield through the local relay). Before anything paid, the viewer sees the cost in ₹ and confirms; a bot or a link can never approve spending. In the editor, agents also get generate workflows (list_gen_workflows, run_gen_workflow), improve_prompt and get_gen_history.

Keys never reach IndiaDigital’s servers. Step-by-step: stock and AI media guide.

Hosting photos and clips

The host must allow other sites to read the file (CORS). These work: raw.githubusercontent.com, res.cloudinary.com, i.imgur.com, upload.wikimedia.org, images.unsplash.com, images.pexels.com / videos.pexels.com, cdn.pixabay.com, dl.dropboxusercontent.com. These don’t: drive.google.com, photos.google.com, instagram.com, facebook.com, whatsapp.net, youtube.com, youtu.be, onedrive.live.com, icloud.com. When a file fails, the page names it and says why; the video is still made from the rest.

Limits and privacy

• Free, no sign-up, no watermark (the “Made with IndiaDigital.org” end-card credit can be turned off with "credit": false).
• Recipe up to 64 KB; up to 30 media files of up to 500 MB each; videos of about 3 minutes. Speed depends on the device that opens the link.
• The recipe sits after the #, so it is never sent to our server or logged. Media go from their host straight to the browser. Nothing is stored.
• The remote MCP is read-only and rate-limited (60 requests a minute per network); it never receives media.

Questions

Can ChatGPT, Grok or Meta AI make a video with IndiaDigital?

Yes. Any assistant that can write a link can make a video: it writes a short recipe into a /video/make/ link and the video renders in the browser that opens it. Assistants that support MCP can call build_video_link on https://indiadigital.org/mcp instead of writing the link by hand.

Is my video or my photos uploaded to IndiaDigital?

No. The recipe travels after the # in the link, which browsers never send to servers. Photos are fetched from their own links straight into the browser, and the video is rendered there. Nothing is stored.

Why does my photo link fail?

The site hosting the photo must allow other sites to read it (CORS). GitHub raw links, Cloudinary, Imgur, Wikimedia, Unsplash, Pexels, Pixabay and Dropbox direct links work; Google Drive, Instagram, WhatsApp and YouTube links do not. The page lists each file that failed and why.

Can a bot edit my own footage with all the tools?

Yes, with the local bridge: run npx indiadigital-video-mcp and connect Claude Code, Cursor, Codex or Gemini CLI. It drives the editor in your own browser tab, footage stays on your computer.

Machine-readable: /video/for-bots/llms.txt · recipe schema · edit-plan schema · all agent tools