Tts

Text-to-Speech

Tool Priority

⚠️ macOS say -o hangs indefinitely for Korean text. Always use edge-tts for Korean file output.

Quick Start (edge-tts)

Install

pip install edge-tts  # In your project venv
brew install ffmpeg   # For format conversion

Generate Voice File

Write a Python script (heredoc/inline causes encoding issues with non-ASCII text):

# /tmp/tts_gen.py
import asyncio, edge_tts

async def main():
    text = "Hello! This is a test voice message from Jaw agent."
    voice = "en-US-JennyNeural"  # or "ko-KR-SunHiNeural" for Korean
    comm = edge_tts.Communicate(text, voice)
    await comm.save("/tmp/voice.mp3")

asyncio.run(main())

python3 /tmp/tts_gen.py

Send to Telegram as Voice Message

# Convert to OGG Opus (Telegram voice format)
ffmpeg -y -i /tmp/voice.mp3 -c:a libopus /tmp/voice.ogg

# Send via Bot API (use python requests — curl multipart may hang)
python3 /tmp/tg_voice.py

# /tmp/tg_voice.py
import json, os, requests

with open(os.path.expanduser("~/.cli-jaw/settings.json")) as f:
    s = json.load(f)
token = s["telegram"]["token"]
chat_id = s["telegram"]["allowedChatIds"][-1]

with open("/tmp/voice.ogg", "rb") as f:
    r = requests.post(
        f"https://api.telegram.org/bot{token}/sendVoice",
        data={"chat_id": chat_id, "caption": "Voice message"},
        files={"voice": f}
    )
print(r.status_code, r.json().get("ok"))

Available Voices

Korean (Recommended)

English

Japanese

List All Voices

edge-tts --list-voices                    # All voices
edge-tts --list-voices | grep ko-KR      # Korean only
edge-tts --list-voices | grep en-US      # US English only

Advanced Options

Speed and Pitch Control

comm = edge_tts.Communicate(
    text,
    "ko-KR-SunHiNeural",
    rate="+20%",     # Speed: -50% to +100%
    pitch="+5Hz",    # Pitch adjustment
    volume="+0%"     # Volume adjustment
)

SSML for Fine Control

ssml = """
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="en-US">
  <voice name="en-US-JennyNeural">
    <prosody rate="medium" pitch="medium">
      Hello there. <break time="500ms"/> Pausing briefly before continuing.
    </prosody>
  </voice>
</speak>
"""

Batch Generation

import asyncio, edge_tts

async def generate(text, output, voice="en-US-JennyNeural"):
    comm = edge_tts.Communicate(text, voice)
    await comm.save(output)

async def main():
    tasks = [
        generate("First message", "/tmp/msg1.mp3"),
        generate("Second message", "/tmp/msg2.mp3"),
        generate("Third in Korean", "/tmp/msg3.mp3", "ko-KR-SunHiNeural"),
    ]
    await asyncio.gather(*tasks)

asyncio.run(main())

macOS say (Fallback — English Only)

say "Hello world"                          # Speak through speakers
say -v Samantha "Hello world"              # Specific voice
say -r 200 "Fast speech"                   # Speed control
say -o /tmp/out.aiff "Hello world"         # Save to file (English OK)
say -v '?' | grep en_                      # List English voices

⚠️ macOS say -o hangs indefinitely for Korean/CJK text. Use edge-tts for non-ASCII file output.

Troubleshooting

Complete Workflow Example

# 1. Generate → 2. Convert → 3. Send → 4. Cleanup
python3 /tmp/tts_gen.py
ffmpeg -y -i /tmp/voice.mp3 -c:a libopus /tmp/voice.ogg
python3 /tmp/tg_voice.py
rm /tmp/tts_gen.py /tmp/tg_voice.py /tmp/voice.mp3 /tmp/voice.ogg