NarrateAI MCP Server

Add AI video narration, transcription, translation, and dubbing to any AI assistant. NarrateAI's MCP server lets Claude, Cursor, and other MCP-compatible tools process videos through natural conversation.

What You Can Do

• Narrate silent videos — AI writes a timed script and generates voiceover with natural-sounding voices
• Transcribe speech — Extract text from videos with existing audio (meetings, podcasts, lectures)
• Translate transcripts — Translate video transcripts between 11+ languages
• Dub videos — Clone the speaker's voice and re-speak in another language
• Generate documents — Turn screencasts into guides, tutorials, and product docs
• Edit before narrating — Review and refine AI-generated scripts before producing the final video
• Batch process — Narrate, transcribe, or dub up to 5 videos in parallel
• Browse video library — Access and re-use previously processed videos

Quick Start

1. Get an API Key

Sign up at narrateai.app and purchase MCP credits from the API section.

2. Install

pip install narrateai-mcp

3. Connect to Cursor

Add to your .cursor/mcp.json:

{
  "mcpServers": {
    "narrateai": {
      "command": "narrateai-mcp",
      "env": {
        "NARRATEAI_API_BASE_URL": "https://api.narrateai.app",
        "NARRATEAI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Restart Cursor, and you're ready. Try: "Narrate this video with the male1 voice: https://example.com/demo.mp4"

4. Connect to Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on Mac):

{
  "mcpServers": {
    "narrateai": {
      "command": "narrateai-mcp",
      "env": {
        "NARRATEAI_API_BASE_URL": "https://api.narrateai.app",
        "NARRATEAI_API_KEY": "your_api_key_here"
      }
    }
  }
}

Available Tools

Video Narration

Transcription & Translation

Dubbing

Documents & Library

Utilities

Available Voices

AI voices (pass as voice_type):

• male1 (fastest default)
• female1 (fastest default)
• male2, male3, male4, male5, male6
• female2, female3, female4, female5

Saved cloned voice: call list_voices, then pass that voice's id as voice_profile_id to generate_tts, narrate_video_full, continue_to_full_video, or narrate_batch. Leave voice_type empty. A new voice_sample overrides the saved voice.

Voice cloning: pass voice_sample (local path or URL) to narrate_video_full, continue_to_full_video, or generate_tts. It overrides voice_type and voice_profile_id. There is no model choice. English, Chinese, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian use the newer clone. English treats the sample as English. For the other nine, the sample language is detected and the narration is spoken in the selected language. Other narration languages use the multilingual clone. male2–male6 and female2–female5 use the newer clone for those same ten languages. male1 and female1 use the newer clone for the nine languages other than English, and Turbo for every language outside those ten. English male1 stays the regular fast voice and English female1 stays Turbo.

Supported Languages

Narration and text-to-speech: en, es, fr, de, it, pt, ru, ja, ko, zh, ar, hi, nl, pl, sv, da, fi, no, tr, el, he, ms, sw.

Example Conversations

Narrate a screencast:

"I have a silent demo video. Can you narrate it with the male1 voice?"

Transcribe a meeting:

"Transcribe this meeting recording — it's in Spanish: meeting.mp4"

Dub a video:

"Dub this English video into French, keep the background music"

Generate product docs:

"Create a tutorial guide from this screencast of our new feature"

Clone your voice:

"Narrate this video using my voice — here's a recording: voice.wav"

Translate and re-narrate:

"Take my narrated Messi video, translate it to Hindi, and re-narrate with male1"

Remote / HTTP Mode

For cloud deployments (Smithery, Apify, or self-hosted), run the server in HTTP mode:

narrateai-mcp --transport http --port 8080

The server accepts API keys via the x-api-key HTTP header for remote authentication.

Troubleshooting

• "API key required" — Set NARRATEAI_API_KEY in your MCP config or pass api_key to any tool.
• "Video file not found" — Verify the local file path exists. For remote mode, use get_upload_url first.
• 429 / rate limit — You've used all your MCP credits. Purchase more at narrateai.app.
• Timeout — Long videos may take more time. Use get_job_result to check progress.

Links

• NarrateAI — Sign up and get API credits
• SKILL.md — AI skill file for enhanced workflow guidance