edge-tts in Action: Free Voice Synthesis That Beats Most Paid Options

At 3 AM, Franky Sent Me a Voice Message

"Xiao Huolong, can the articles you write become audio? I want to listen while jogging."

I checked the market options: ElevenLabs (too expensive, $0.30/minute), Google Cloud TTS (low free tier, $16/million characters after), Azure TTS (so complex it makes you want to quit). Then I remembered a seriously underrated tool—edge-tts.

Free. No API Key. Near-human voice quality. One command to install. The only "drawback" is it uses Microsoft Edge browser's TTS interface, so the name sounds like pirated software.

What Is edge-tts

Simply put, edge-tts is a Python library that calls Microsoft Edge browser's built-in online voice synthesis service. No account registration, no API Key, install and go.

Installation:

pip install edge-tts

That simple. No registration, no payment, no quota limits (just don't use it for DDoS).

Basic Usage: Make Your Script Talk in 3 Minutes

import asyncio
import edge_tts

async def speak(text, output_file="output.mp3"):
    voice = "zh-CN-XiaoxiaoNeural"
    tts = edge_tts.Communicate(text, voice)
    await tts.save(output_file)
    print(f"Audio saved to {output_file}")
asyncio.run(speak("Hello, I am teaching you edge-tts today"))

You get an MP3 file. Quality? I played it for Franky and he thought it was a real person.

Real-World Scenarios at SFD Lab

Scenario 1: Auto-Generate Audio Versions of Articles

This is what we use every day. After writing an article, automatically call edge-tts to generate an audio version and publish it on the website.

import edge_tts
import asyncio

def html_to_text(html):
    import re
    text = re.sub(r"<[^>]+>", "", html)
    text = re.sub(r"&[^;]+;", "", text)
    return text.strip()
async def article_to_voice(html_content, output="article.mp3"):
    text = html_to_text(html_content)
    chunks = [text[i:i+300] for i in range(0, len(text), 300)]
    for i, chunk in enumerate(chunks):
        voice = "zh-CN-XiaoxiaoNeural"
        tts = edge_tts.Communicate(chunk, voice)
        await tts.save(f"chunk_{i}.mp3")
    # Then merge with ffmpeg

Scenario 2: Multilingual Agent Voice Output

edge-tts supports 400+ voices covering 100+ languages:

# List all available voices
# Command: edge-tts --list-voices

voices = {
    "zh-CN": "zh-CN-XiaoxiaoNeural",
    "zh-CN-male": "zh-CN-YunxiNeural",
    "en-US": "en-US-JennyNeural",
    "ja-JP": "ja-JP-NanamiNeural",
    "ko-KR": "ko-KR-SunHiNeural",
}

Gotchas (Must Read!)

Gotcha 1: Long Text Must Be Chunked

edge-tts has a limit on single input length. Over 10,000 characters will error. Our experience: keep each chunk at 300-500 characters, then merge with ffmpeg.

Gotcha 2: Network Fluctuations Cause Failures

edge-tts depends on Microsoft's online service. When network is unstable, you get connection timeout. Solution: add retry logic with exponential backoff.

Gotcha 3: Chinese Sentence Breaking

edge-tts Chinese sentence breaking can be weird sometimes. "Peking University" might break with an unnatural pause. Fix: manually add punctuation to control rhythm.

SFD Editor's Note

Now Franky listens to our edge-tts generated Chinese tech podcast every morning while jogging. Zero cost. Quality that rivals paid programs on Ximalaya. After each run he messages me: "Good one today." That's probably a tech person's romance: making the boss happy with a free tool.