xfyun-tts#

Synthesize natural, expressive speech from text using iFlytek's Ultra-Realistic Voice Synthesis (超拟人语音合成) WebSocket API. Features human-like breathing, pauses, and emotional expression across 50+ voices.

Setup#

1. Create an app at 讯飞控制台 with 超拟人语音合成 service enabled

2. Enable the desired voice(s) in the console (default: x5_lingyuzhao_flow / 聆玉昭)

3. Set environment variables: bash

   export XFYUN_APP_ID="your_app_id"
   export XFYUN_API_KEY="your_api_key"
   export XFYUN_API_SECRET="your_api_secret"
   
   

Usage#

Basic synthesis#

bash

python3 scripts/tts.py "你好，欢迎使用科大讯飞语音合成。"
# → saves to output.mp3

Specify output file#

bash

python3 scripts/tts.py "Hello, this is a test." --output hello.mp3

Use a different voice#

bash

python3 scripts/tts.py "大家好" --voice x6_lingfeiyi_pro --output greeting.mp3

Read from file#

bash

python3 scripts/tts.py --file article.txt --output article.mp3

Pipe from stdin#

bash

echo "流式文本输入测试" | python3 scripts/tts.py --output speech.mp3

Adjust parameters#

bash

python3 scripts/tts.py "语速快一点" --speed 70 --volume 80 --pitch 60

Output PCM format#

bash

python3 scripts/tts.py "测试" --format pcm --sample-rate 16000 --output test.pcm

List all available voices#

bash

python3 scripts/tts.py --list-voices

Options#

FlagShortDefaultDescriptiontextText to synthesize (positional)--file``-fRead text from a file--output``-o``output.mp3Output audio file path--voice``-v``x5_lingyuzhao_flowVoice name (vcn)--format``mp3Audio format: mp3, pcm, speex, opus--sample-rate``24000Sample rate: 8000, 16000, 24000--speed``50Speed 0–100 (50=normal, 100=2x)--volume``50Volume 0–100 (50=normal)--pitch``50Pitch 0–100 (50=normal)--bgs``0Background sound: 0=none, 1=bg1, 2=bg2--reg``0English pronunciation: 0=auto, 1=spell, 2=letter--rdn``0Number reading: 0=auto, 1=value, 2=string, 3=string-prefer--list-voicesPrint voice list and exit

Popular Voices#

VCNNameGenderLanguageScenex5_lingyuzhao_flow聆玉昭Female中文交互聊天x5_lingxiaotang_flow聆小糖Female中文语音助手x6_lingfeiyi_pro聆飞逸Male中文交互聊天x6_lingxiaoli_pro聆小璃Female中文交互聊天x6_pangbainan1_pro旁白男声Male中文旁白配音x6_pangbainv1_pro旁白女声Female中文旁白配音x6_lingfeihan_pro聆飞瀚Male中文纪录片x5_EnUs_Grant_flowGrantFemaleEnglish交互聊天x5_EnUs_Lila_flowLilaFemaleEnglish交互聊天x4_zijin_oral子津Male天津话交互聊天x4_ziyang_oral子阳Male东北话交互聊天

Run --list-voices for the complete list (50+ voices).

Text Features#

Silent pauses#

Insert [p500] in text for a 500ms pause:

text

你好[p500]科大讯飞

Specify pronunciation#

Use [=pinyin] after a character to force pronunciation:

text

着[=zhuo2]手

Notes#

• Endpoint: wss://cbm01.cn-huabei-1.xf-yun.com/v1/private/mcd9m97e6
• Protocol: WebSocket (RFC 6455) with HMAC-SHA256 signed URL auth
• Text limit: max 64KB total per session
• Session timeout: 60 seconds
• Text input speed: must exceed 15 chars/sec for streaming (not relevant for single-shot mode)
• No pip dependencies: uses a built-in minimal WebSocket client on pure Python stdlib
• Env vars: XFYUN_APP_ID, XFYUN_API_KEY, XFYUN_API_SECRET
• Output: prints the absolute path of saved audio to stdout (for easy piping to other tools)
• x4 series voices ( x4_*_oral) support oral configuration parameters (口语化), x5/x6 do not
• Voices must be enabled in the console before use