evo-tts-audiobook-synth

Converts text chunks into speech audio using cascading fallback (OpenAI TTS -> ElevenLabs -> edge-tts -> gTTS). Handles API auth, rate limiting, MP3 chunk generation, and audio concatenation into a si

SKILL.md

GitHub

evo-tts-audiobook-synth

Converts text to speech with cascading API fallback and audio concatenation.

Key Functions

  • synthesize_with_openai(text, output_path) - Uses OpenAI TTS API (tts-1, 4096 char limit)
  • synthesize_with_elevenlabs(text, output_path) - Uses ElevenLabs API (client.text_to_speech.convert)
  • synthesize_with_local_tts(text, output_path) - Uses edge-tts (async) or gTTS as fallback
  • synthesize_text_to_audio(text, output_path) - Tries OpenAI -> ElevenLabs -> local, returns bytes
  • concatenate_audio_chunks(chunk_files, output_file) - Concatenates MP3 files using pydub/ffmpeg

Usage

import sys
sys.path.insert(0, '/app/environment/skills/evo-tts-audiobook-synth/scripts')
from utils import synthesize_text_to_audio, concatenate_audio_chunks

# Synthesize each chunk
chunk_files = []
for i, chunk in enumerate(chunks):
    path = f'/tmp/chunk_{i}.mp3'
    synthesize_text_to_audio(chunk, path)
    chunk_files.append(path)

# Concatenate all chunks
concatenate_audio_chunks(chunk_files, '/root/audiobook.mp3')

Technical Notes

  • OpenAI TTS: client.audio.speech.create(model="tts-1", voice="onyx", input=text), max 4096 chars
  • ElevenLabs: client.text_to_speech.convert(text=text, voice_id="...", model_id="eleven_multilingual_v2")
  • edge-tts: async, uses Microsoft Edge neural voices, no API key needed
  • gTTS: Google Translate TTS, no API key needed, lower quality
  • Audio concat: use pydub AudioSegment or ffmpeg subprocess
  • Always use "wb" mode for writing audio binary data
  • Add time.sleep(1) between API calls to avoid rate limiting