Zum Inhalt springen
TakoAPI
Zurück zu den Skills

sergei-mikhailov-stt

von bzsega

Communication
441Downloadzahlen laut clawskills.sh
0Aufrufe
0 LikesMelde dich an, um diesen Skill zu liken
—Noch nicht bewertet
Deine Bewertung

Kurzbeschreibung

Speech recognition from voice messages.

Beschreibung

Was dieser Skill macht

Converts voice messages received through any OpenClaw-connected messenger into text using Yandex SpeechKit. Supports OGG, WAV, MP3, M4A, FLAC, and AAC formats up to 1 MB (~30 seconds of audio). The provider architecture is extensible, so alternative STT services can be added without modifying the core processor.

Runs directly within OpenClaw without requiring a separate transcription service or manual file upload workflow.

Wann du ihn einsetzt

  • Transcribing Telegram voice messages into readable text
  • Getting text versions of audio replies when playback is not possible
  • Logging spoken notes sent through a connected messenger bot
  • Archiving voice messages as searchable text records
  • Converting Russian or English voice input from any connected chat

Beispiel-Workflow

EINGABE

User asks: [sends a voice message via a connected messenger]

AGENT
  1. 1

    Receives the local audio file path from OpenClaw

  2. 2

    Validates file format and size against Yandex SpeechKit limits

  3. 3

    Converts audio to a supported format using ffmpeg if needed

  4. 4

    Sends the audio file to Yandex SpeechKit and retrieves the result

  5. 5

    Returns the transcribed text with detected language and confidence score

AUSGABE

Recognized text: "Meeting tomorrow at 3 PM" (ru-RU, 95% confidence, provider: Yandex SpeechKit)

Voraussetzungen

  • Yandex Cloud account with a service account that has the ai.speechkit.user role
  • Yandex SpeechKit API key (YANDEX_API_KEY) generated from Yandex Cloud IAM
  • Yandex Cloud folder ID (YANDEX_FOLDER_ID) from the Yandex Cloud console

Installation

clawhub install bzsega/sergei-mikhailov-stt

Führe den CLI-Befehl aus oder füge den Prompt in Claude Code, Cursor, Codex oder deinen Coding-Agenten ein.

Links