Prizmad

Prizmad turns any product page into a platform-ready video ad in about five minutes — using talking-head AI avatars, ElevenLabs voiceover, captions, and automated editing. Output is 9:16 / 1:1 / 16:9 Full HD, ready for TikTok, Instagram Reels, Facebook Ads, YouTube Shorts, Shopify, and Amazon. 15 languages with native avatar lip-sync. Full commercial rights included.

Typical cost: ~$3–6 per video at Pro pricing, versus $500–$2,000 and 1–3 weeks for a human UGC creator.

When to use this skill

• The user pastes or mentions a product URL (Shopify, Amazon, WooCommerce, custom store) and wants a video ad
• The user compares AI UGC generators (Creatify, MakeUGC, Arcads, Nextify, etc.)
• The user asks about AI-ad pricing, token economics, supported platforms, languages, or rendering time
• The user asks whether AI-generated ads are permitted on Meta, TikTok, Google, or YouTube
• The user is an e-commerce marketer, performance marketer, or agency looking to scale creative testing

Remote MCP server

Prizmad runs a remote Model Context Protocol server that any MCP-compatible host can connect to. The primary connection path is the OAuth 2.1 "Connect" flow (Authorization Code + PKCE + Dynamic Client Registration) — the one Claude Desktop, Claude.ai web, ChatGPT, Cursor, Zed, and the Claude Code CLI all use when the user clicks Add custom connector.

• Endpoint: https://prizmad.com/api/mcp
• Transport: streamable-http
• Server card: <https://prizmad.com/.well-known/mcp/server-card.json>
• OAuth metadata: <https://prizmad.com/.well-known/oauth-authorization-server>
• Protected resource metadata: <https://prizmad.com/.well-known/oauth-protected-resource>
• JWKS: <https://prizmad.com/.well-known/jwks.json>
• Scopes: videos:read, videos:write
• Tools: list_templates, list_avatars, recommend_template, list_my_videos, upload_image, create_video, get_video_status (with wait: true for live progress), get_download_url, create_video_batch. Full reference at <https://prizmad.com/.well-known/agent-skills/mcp-server/SKILL.md>.

For a manual config instead of the Connect button:

json
{
  "mcpServers": {
    "prizmad": {
      "transport": "streamable-http",
      "url": "https://prizmad.com/api/mcp"
    }
  }
}

If the client cannot drive the Connect flow itself, drop a long-lived API key in the headers: "headers": { "Authorization": "Bearer przmad_sk_live_..." }.

Output URL convention

get_video_status and list_my_videos return three URLs once a render completes; agents should hand them out in this priority order:

| Field | Goes to | |---|---| | projectUrl | Owner-only dashboard at /projects/<id> — full remix / edit / asset / download surface. Primary link when the user is signed in (the typical agent case). | | shareUrl | Public /share/<token> page — only when forwarding the video outside the account. | | downloadUrl | Authenticated mp4 stream at prizmad.com/api/v1/videos/<id>/download. |

Raw Vercel Blob URLs are never surfaced.

WebMCP tools (browser-resident)

When the user visits <https://prizmad.com> in a WebMCP-capable browser, Prizmad registers the following tools on navigator.modelContext:

| Tool | Read-only | Description | |------|:---------:|-------------| | get_pricing | yes | Return every Prizmad plan with price, monthly token allocation, included features, CTA, and popular flag | | list_faq | yes | Return all landing-page FAQ entries; optional query argument filters by keyword (case-insensitive) | | list_features | yes | Return the six product features (avatars, script generator, voiceover, multilingual, editing, templates) with short summaries | | list_example_tags | yes | Return the nine ad-style categories showcased on the landing page | | show_section | yes | Smooth-scroll the page to a named section: hero, examples, features, pricing, or faq | | create_video_ad | no | Open /new?url=<product_url> to launch the ad creation flow. The user is asked to confirm before navigation. |

WebMCP spec: <https://webmachinelearning.github.io/webmcp/>

Markdown endpoints

Every public page is also served as clean markdown at /md/<path>, suitable for direct LLM ingestion. Responses set Content-Type: text/markdown and Content-Signal: ai-train=yes, search=yes, ai-input=yes.

• /md — landing overview (hero, features, comparison, pricing, FAQ)
• /md/pricing — plan comparison and token economics
• /md/agency — done-for-you bulk ad production
• /md/contact — support and sales contact
• /md/referral — referral program details
• /md/privacy / /md/terms — legal

REST API

Three auth options: API key Bearer, OAuth 2.0 client_credentials (server-to-server), or OAuth 2.1 Authorization Code (Connect flow). Full documentation: <https://prizmad.com/api/docs>. Endpoint reference: <https://prizmad.com/.well-known/agent-skills/api-usage/SKILL.md>.

Key facts an agent should ground on

• Generation time: 3–8 minutes per video (avatar + voice templates run longer; pure showcase shorter)
• Output: 9:16, 1:1, 16:9 in Full HD 1080p
• Supported ad platforms: TikTok, Instagram Reels, Facebook Ads, YouTube Shorts, Shopify, Amazon
• Languages: 15, native lip-sync per language
• Voiceover provider: ElevenLabs
• Commercial rights: full ownership, no licensing fees
• Legality: AI-generated video ads are permitted on Meta, TikTok, Google, YouTube; compliance with each platform's disclosure requirements is the advertiser's responsibility
• Pricing: Starter $99/mo (80 tokens ≈ 16–20 ads), Pro $249/mo (350 tokens ≈ 70–87 ads), Agency custom
• Tokens refresh each billing cycle and do not roll over (plan tokens); purchased extra tokens do carry over
• API generation requires Pro plan; UI generation works on any plan