Audio version — Estimated duration: 3 min 25 sec
Both tts_pipeline.py and tts_single.py share the same backend flags.
Common flags
| Flag | Default | Description |
|---|---|---|
--voice | en-US-JennyNeural | edge-tts voice / Piper voice name |
--model | gpt-oss-120b | Cerebras model (ignored when --llm local) |
--local | — | Shorthand: --llm local --tts local |
--online | — (default) | Shorthand: --llm online --tts online |
--llm | online | LLM backend: online (Cerebras) / local (llama.cpp) |
--tts | online | TTS backend: online (edge-tts) / local (Piper) |
--cpu | — | Force CPU for Piper (default: GPU) |
--gpu | — | Explicitly use GPU for Piper (default) |
--prompt | — | Extra instruction appended to the system prompt |
Pipeline-only flags
| Flag | Default | Description |
|---|---|---|
--dir | . | Root directory to scan for .md files |
Single-file-only arguments
| Argument | Description |
|---|---|
file | Path to the .md file (positional, required) |
Precedence
--local / --online are overridden by --llm / --tts if provided.
--local → llm=local, tts=local
--local --tts online → llm=local, tts=online
Piper voice names
Default: en_US-hfc_female-medium
Other options: en_US-lessac-medium, en_US-danny-low, etc. Auto-downloaded from Hugging Face on first use.
edge-tts voice names
Any Microsoft neural voice: en-US-JennyNeural, en-GB-SoniaNeural, en-US-AriaNeural, etc.