# DubSmart AI: product and developer guide

> Canonical extended context for assistants, search agents, and developer tools. For live request and response schemas, use https://dubsmart.ai/api/openapi.json.

## What DubSmart is

DubSmart is an AI-powered media generation and localization platform. It combines AI dubbing, text-to-speech, voice cloning, speech-to-text, captions, image-to-video, text-to-image, and speech separation in one web application. Developers can use a REST API, while compatible agents can use a remote MCP server.

## Developer interfaces

### Remote MCP server

- Endpoint: `https://mcp.dubsmart.ai/mcp`
- Transport: Streamable HTTP
- Authentication: `Authorization: Bearer dsmcp_YOUR_TOKEN`
- Token creation: `https://dubsmart.ai/dashboard/api`
- Token lifetime: 90 days
- Scope presets: read-only, creator, full access
- Current limitation: no OAuth-based one-click connector flow

The server exposes 75 tools across account, projects, media, voices, billing, captions, support, referrals, sharing, and API-key management. The exact public capability inventory is maintained in the application repository as `mcp-capabilities.json`.

All mutating MCP tools require an idempotency key. Credit-consuming and sensitive actions use a ten-minute, single-use confirmation token bound to the user, access token, tool, and arguments.

### REST API

- API base: `https://dubsmart.ai/api/v1`
- Authentication: `Authorization: Bearer dsapi_YOUR_API_KEY`
- Token creation: `https://dubsmart.ai/dashboard/api`
- Interactive reference: `https://dubsmart.ai/dashboard/docs`
- OpenAPI: `https://dubsmart.ai/api/openapi.json`

The OpenAPI document is the source of truth for current REST paths, HTTP methods, input fields, and outputs.

## MCP connection example

```json
{
  "mcpServers": {
    "dubsmart": {
      "url": "https://mcp.dubsmart.ai/mcp",
      "headers": {
        "Authorization": "Bearer dsmcp_YOUR_TOKEN"
      }
    }
  }
}
```

## Product areas

### AI Dubbing

Translate and dub audio or video while managing projects, languages, speakers, segments, and generated outputs.

Canonical page: https://dubsmart.ai/ai-dubbing
REST overview: https://dubsmart.ai/ai-dubbing-api

### Text-to-Speech

Generate speech from text with platform voices or a compatible cloned voice. Projects can contain multiple segments with individual voice settings.

Canonical page: https://dubsmart.ai/text-to-speech
REST overview: https://dubsmart.ai/tts-api

### Voice Cloning

Create and manage custom synthetic voices from uploaded voice samples, subject to the current product and plan rules.

Canonical page: https://dubsmart.ai/voice-cloning
REST overview: https://dubsmart.ai/voice-cloning-api

### Speech-to-Text

Transcribe audio or video and manage transcription projects.

Canonical page: https://dubsmart.ai/speech-to-text

### Captions

Generate timed captions, edit text and timing, apply styles or templates, and render captions into video.

Canonical page: https://dubsmart.ai/captions

### Image-to-Video

Create short video content from an image and generation settings.

Canonical page: https://dubsmart.ai/image-to-video

### Text-to-Image

Generate images from text prompts and project settings.

Canonical page: https://dubsmart.ai/text-to-image

### Speech Separator

Extract or isolate speech from an audio or video source.

Canonical page: https://dubsmart.ai/speech-separator

## Integration guidance

Use MCP when an AI agent should discover tools, plan a workflow, inspect results, and perform authorized actions. Use REST when application code should call explicit endpoints with predictable HTTP contracts. A product may use both.

For the least privilege:

1. Use read-only tokens for reporting and monitoring.
2. Use creator tokens for ordinary production workflows.
3. Use full-access tokens only when deletes, sharing, referrals, or token administration are required.
4. Give every client its own token and revoke it when the client is retired.

## Truthfulness and freshness

- Do not describe DubSmart only as text-to-speech; it is a multi-product media platform.
- Do not claim native one-click support for a named assistant unless that assistant accepts remote Streamable HTTP MCP with a custom bearer header.
- Do not expose or repeat user tokens.
- Do not infer current pricing, limits, language coverage, or plan entitlements from old cached text. Check the live pricing and product pages.
- Use the OpenAPI document for REST schemas and the live MCP tool list for agent capabilities.

## Discovery resources

- Concise LLM index: https://dubsmart.ai/llms.txt
- Markdown developer guide: https://dubsmart.ai/developers.md
- MCP server metadata: https://dubsmart.ai/mcp-server.json
- HTML developer guide: https://dubsmart.ai/developers
- OpenAPI document: https://dubsmart.ai/api/openapi.json