Audio version — Estimated duration: 1 min 33 sec

Convert markdown documentation into spoken audio with independent LLM and TTS backends — mix and match online and local engines.


Features

  • Dual backends — Cerebras API or llama.cpp for LLM; edge-tts or Piper for TTS
  • Batch mode — Recursively process an entire .md directory tree
  • Single mode — Convert one file with a single command
  • Audio embedding — Injects <audio> tags back into source files with duration metadata
  • Conversational output — LLM rewrites docs into spoken-word text before synthesis

Backend matrix

ComponentOnlineLocal
LLMCerebras APIllama.cpp server (localhost:8080)
TTSedge-tts (Microsoft cloud)Piper (GPU via onnxruntime)

Quick nav

SectionWhat’s there
Installationpip install, API key, local server setup
QuickstartGenerate your first audio in 2 commands
Pipeline ModeBatch-process a directory of .md files
Single File ModeConvert one .md file
CLI ReferenceEvery argument explained
BackendsOnline and local LLM/TTS setup guides
OutputFile layout, formats, audio embedding
ArchitecturePipeline internals, rate limiting, ordering
TroubleshootingCommon errors and fixes

Tech stack: Python 3.10+ · OpenAI SDK · edge-tts · Piper · onnxruntime-gpu · Cerebras API · llama.cpp

4 items under this folder.