Supported Providers

AudioBard supports a variety of local offline backends and cloud BYOK (Bring Your Own Key) providers.

---

LLM Providers (Character Extraction & Attribution)

Provider	Type	Privacy	Recommended Models	Environment Variable
Ollama	Local / Offline	100% Private	qwen2.5:7b, llama3.3:70b	None (Local host)
Google Gemini	Cloud BYOK	Cloud	gemini-2.0-flash, gemini-1.5-pro	GEMINI_API_KEY
NVIDIA NIM	Cloud BYOK	Cloud	meta/llama-3.3-70b-instruct	NVIDIA_API_KEY
OpenRouter	Cloud BYOK	Cloud	anthropic/claude-3.5-haiku	OPENROUTER_API_KEY

Configuring Ollama (Offline Default)

Ensure Ollama is running locally:

ollama run qwen2.5:7b

Configuring Cloud API Keys

Add your keys to a .env file in the repository root:

GEMINI_API_KEY=your_gemini_api_key_here
NVIDIA_API_KEY=your_nvidia_api_key_here
OPENROUTER_API_KEY=your_openrouter_api_key_here

---

TTS Providers (Voice Synthesis)

Provider	Type	Privacy	Highlights
Piper TTS	Local / Offline	100% Private	Ultra-fast CPU neural synthesis, zero latency, lightweight ONNX models.
Edge TTS	Cloud (Free)	Cloud	High-fidelity neural voices with diverse regional accents and prosody support.

---

Configuration File (config.yaml)

You can persist your default provider selections in config.yaml:

llm:
  provider: ollama        # ollama | gemini | openrouter | nim
  model: qwen2.5:7b
  temperature: 0.1

tts:
  provider: piper         # piper | edge
  locale: en_US

audio:
  format: mp3             # mp3 | m4b
  bitrate: 192k
  silence_gap_ms: 300
  normalize_volume: true