# transcribe-mcp [Health: Active]

**Category:** 🎙️ Speech-to-Text  
**Repository:** https://github.com/Soboljem/transcribe-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 365  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/transcribe-mcp

## Description
Transcribe audio/video and find dead air. Pay-per-call in USDC over x402 — no account, no key.

## Tools
Capabilities this server exposes over MCP:

- **transcribe_url** — a public audio/video `url` **or a local audio `file`** → text, timings, detected language
- **find_dead_air** — subtitles or timings (inline, or `from_file` a .json) → the spans worth cutting, and what you'd save
- **fetch_page** — a web page → clean markdown, links resolved. `render: true` for JavaScript pages
- **shop_info** — prices, limits and how payment works

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "transcribe-mcp": {
    "command": "npx",
    "args": ["-y","transcribe-mcp"],
    "env": {
      "TRANSCRIBE_MCP_WALLET_KEY": ""
    }
  }
}
```

**Requires environment variables:** `TRANSCRIBE_MCP_WALLET_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What transcribe-mcp MCP server does

The transcribe-mcp MCP server provides MCP tools for two related media-processing tasks: obtaining a transcript from a public audio or video URL, and locating silent sections that may be removed during editing. The transcription result includes text, timing information, and a detected language. The dead-air analysis accepts subtitles or timings and returns candidate cut ranges along with the duration and percentage that could be saved.

The tools are designed to work together. An agent can first call `transcribe_url`, then pass its timing or subtitle output to `find_dead_air`. For pricing, usage limits, and payment details, `shop_info` provides a free discovery call.

## How it works

Billing uses the x402 HTTP payment flow rather than an account-based API. A paid request receives a price requirement, the client pays in USDC, and the request proceeds after payment. Settlement takes place on Base, identified as `eip155:8453`.

The listed prices are $0.05 for `transcribe_url` and $0.002 for `find_dead_air`; `shop_info` does not cost anything. The client applies a $0.10 maximum per call. If a quoted price exceeds that ceiling, the request is rejected instead of being paid. The client also checks for settlement confirmation and does not report a successful payment when the expected settlement header is absent.

## Setup and configuration

A standard MCP client can start the package with `npx`:

```json
{
  "mcpServers": {
    "transcribe": {
      "command": "npx",
      "args": ["-y", "transcribe-mcp"],
      "env": {
        "TRANSCRIBE_MCP_WALLET_KEY": "0x..."
      }
    }
  }
}
```

The transcribe-mcp MCP server does not require an account, subscription, or API key. `TRANSCRIBE_MCP_WALLET_KEY` is needed only for the paid tools and should contain the private key of a wallet holding USDC on Base. A dedicated wallet with a small balance is recommended instead of a primary wallet. Without a wallet, the process can still start, list its tools, and provide `shop_info`; paid calls explain what configuration is missing.

## Tools and capabilities

- `transcribe_url`: Accepts a public audio or video URL and returns transcript text, timings, and detected language. The listed price is $0.05 per call.
- `find_dead_air`: Accepts subtitles or timing data and reports silence ranges, their duration, and estimated savings. The listed price is $0.002 per call.
- `shop_info`: Reports prices, limits, and payment behavior at no charge.

A typical workflow is to transcribe a talk, feed the returned timing data into `find_dead_air`, and use the resulting ranges for editing decisions. The tool output can include cut objects with start and end positions, duration, and a reason such as silence, plus source, speech, and savings statistics.

## Limitations and notes

Inputs are limited to 800 KB. Oversized inputs are refused before payment, so they do not incur a charge. Invalid arguments are also checked locally and do not leave the client. A failed shop response does not settle a payment.

The transcription tool requires a publicly reachable media URL, and the payment wallet must hold USDC on Base for paid requests. The material does not describe private media access, file-upload workflows, editing output, or support for payment networks other than Base.

_Full upstream README: https://allmcps.com/mcp/transcribe-mcp/readme_