A smart Telegram workflow for OpenClaw that provides intelligent text and voice responses using local MLX-based transcription and synthesis.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install telegram-multilingual-voice-reply
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install telegram-multilingual-voice-reply using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Telegram Multilingual Voice Reply?
The Telegram Multilingual Voice Reply skill is a sophisticated communication workflow designed for AI agents. It intelligently differentiates between text and audio inputs to provide the most natural response format. By leveraging local MLX models on Apple Silicon, it ensures high-speed, private transcription and speech generation without relying on external cloud APIs.
This skill is a vital addition to Openclaw Skills for users who want a hands-free, multilingual experience. It follows the user's language automatically—replying in Chinese if the input is Chinese, or English if the input is English—unless a specific override is requested. Its unique ability to bundle voice notes and text captions into a single Telegram message significantly improves the user interface and accessibility.
Telegram Multilingual Voice Reply Use Cases
- Hands-free AI interaction via Telegram voice notes while driving or walking.
- Localized, private transcription of sensitive audio messages on macOS.
- Multilingual customer support or personal assistance with automatic language detection.
- Creating a seamless voice-and-text record of AI conversations within a single chat bubble.
How Telegram Multilingual Voice Reply Works
- The agent detects if the incoming Telegram message is text or a voice/audio file.
- If the input is text, the agent responds with text by default to maintain conversational context.
- If a voice note is received, the skill invokes a local ASR model (typically Qwen3-ASR) via MLX to transcribe the audio.
- The agent generates a contextual reply based on the transcription in the same language as the user.
- The skill then triggers a TTS engine to create a Telegram-compatible .ogg/opus file.
- Finally, it sends the response as a single voice message with the text response included as a caption.
Telegram Multilingual Voice Reply Setup
To use this skill within your Openclaw Skills environment, ensure you have the mlx_audio package installed in your Python environment. You can test the underlying scripts using the following commands:
# To transcribe an audio file locally
python3 scripts/mlx_asr.py --audio /path/to/audio.ogg --language Chinese
# To generate a voice reply for Telegram
python3 scripts/mlx_tts_voice.py --text "Hello, this is your AI assistant." --out /tmp/reply.ogg
Ensure ffmpeg is installed on your system to handle audio format conversions from Telegram's OGG format to WAV for processing.
Telegram Multilingual Voice Reply Data Schema & Taxonomy
The skill utilizes the OpenClaw message tool with specific parameters to ensure voice and text parity:
| Parameter | Requirement | Description |
|---|---|---|
| asVoice | true | Forces the message to be sent as a native Telegram voice note. |
| filePath | string | The local path to the generated .ogg or .opus file. |
| caption | string | The text content of the reply, which must match the audio. |
| language | dynamic | Determined by the ASR output to ensure the response matches the input. |
Telegram Multilingual Voice Reply Advanced Features
- Automatic language mirroring for Chinese, English, and other supported languages.
- Local execution on Apple Silicon for low latency and high privacy.
- Smart intent detection to switch between text-only and voice+text responses.
- Support for Qwen3-ForcedAligner to improve transcription accuracy and timing.
- Single-message delivery for voice notes and captions, optimizing the Telegram UI.
SKILL.md
Loading
Related Openclaw Skills
Glucksrad Decision Wheel
An interactive 3D decision-making skill that generates custom spinning wheels for random selection.
TRMNL Content Generator
A professional toolset for generating and deploying dithered HTML layouts to TRMNL e-ink devices using a specialized CSS framework and CLI.
Cal.com API v2 Integration
A powerful integration for AI agents to automate scheduling, manage bookings, and synchronize calendars via the Cal.com API v2.
Local Life
A comprehensive local intelligence dashboard providing real-time weather, financial data, and holiday information for any city.
AI Learning Journal & Growth Tracker
A structured documentation and roadmap tool designed to help developers and enthusiasts systematically track their AI learning journey.
BLE-Anima-Minter
A proximity-based tokenization tool that converts Bluetooth signals into soulbound $ANIMA assets stored in a local DAG.
Featured*
