mcp-transcribe mcp

v1.0.0 · MCP Tool · ai · registry.pascalai.org

Transcribe audio files to text using Whisper or compatible speech-to-text providers. Supports language detection, word-level timestamps, speaker diarization and multiple output formats.

transcribespeech-to-textwhisperaudiovoicestt

MakerAI Pipeline

Input Parameters

ParameterTypeDescription
audio_urloptional string URL of the audio file (mp3, mp4, wav, m4a, ogg, webm).
audio_b64optional string Base64-encoded audio content (alternative to audio_url).
languageoptional string Language code (e.g. 'en', 'es', 'fr'). Empty = auto-detect. Default: .
modeloptional string Whisper model variant. Default: 'auto'. One of: whisper-1, whisper-large-v3, distil-whisper, auto. Default: auto.
timestampsoptional string Timestamp granularity. Default: 'segment'. One of: none, segment, word. Default: segment.
diarizeoptional boolean Attempt speaker diarization (identify different speakers). Default: false. Default: False.
formatoptional string Output format. Default: 'text'. One of: text, json, srt, vtt. Default: text.

Output Fields

FieldTypeDescription
text string Full transcription text.
language string Detected or specified language.
duration_seconds number
segments array[object] Timestamped segments.
formatted string Formatted output (SRT/VTT when format != 'text').
model string

Examples

Transcribe a podcast episode

// Input
{
  "audio_url": "https://example.com/podcast_ep42.mp3",
  "language": "en",
  "timestamps": "segment"
}

// Output
{
  "text": "Welcome to episode 42. Today we discuss...",
  "language": "en",
  "duration_seconds": 2834.5,
  "segments": [
    {
      "start": 0.0,
      "end": 4.2,
      "text": "Welcome to episode 42."
    }
  ]
}

Install & Discovery

Install

ppm install mcp-transcribe

Get JSON Schema

GET /v1/packages/mcp-transcribe/1.0.0/schema

Discover by keyword

GET /v1/mcp/discover?q=transcribe
Discovery hint: Install with ppm install mcp-transcribe or invoke remotely via POST /v1/invoke/mcp-transcribe on the MCP Service.

PascalAI Usage

uses toolslib;
var Tool := LoadTool('mcp-transcribe');
var R := Tool.Call(JsonObj(['audio_url','https://example.com/meeting.mp3','timestamps','word']));
Writeln(R['text']);