Free open-source AI

Be A Buddhist v2 will use $0 AI — no Suno subscription, no OpenAI bills. Self-host on your machine or use Hugging Face's free inference tier for demos.

Ambient · MIT

Meta MusicGen

Text-to-music generation. Prompt: "soft temple bells, 20 minute meditation ambient, no drums" → backing track for a routine. Part of AudioCraft (Meta, MIT license).

# Self-host (GPU recommended)
pip install audiocraft
python -m audiocraft.demo musicgen --prompt "temple bells meditation"

musicgen-small on Hugging Face — free serverless inference (rate-limited, fine for personal use).

SFX · TTS · MIT

Suno Bark (open source)

Text-to-audio: speech, chanting-style vocals, sound effects. github.com/suno-ai/bark — MIT license. Good for short guided intros or bell sounds from text prompts.

pip install git+https://github.com/suno-ai/bark.git
# Generate from prompt string

Guided voice · MIT

Piper TTS

Fast, local text-to-speech. Run offline on phone or server — perfect for "begin your metta practice" intros without cloud costs. github.com/rhasspy/piper

# ONNX models, runs on CPU
echo "Begin your metta practice" | piper --model en_US-lessac-medium

Free tier · API

Hugging Face Inference API

Free tier for open models — MusicGen, Bark, and others. huggingface.co/inference-api Rate limits apply; self-host when you outgrow them.

POST https://api-inference.huggingface.co/models/facebook/musicgen-small
Authorization: Bearer HF_TOKEN  # free account

Ambient · Open weights

Stable Audio Open

Stability AI's open model for short audio clips (≤47s). stable-audio-open-1.0 — useful for transition stingers and short ambient loops.

v2 integration strategy

Use case	Model	Cost
Routine ambient bed	MusicGen small	Free (HF or self-host)
Guided intro voice	Piper TTS	Free, offline
Bells / nature SFX	Bark or Web Audio	Free
Extend chant loop	AudioCraft MusicGen	Free
Production scale	Self-host on GPU VPS	~$5–20/mo optional

Abstraction layer

Phase 3 adds an AudioGenProvider interface:

interface AudioGenProvider {
  generateAmbient(prompt: string, durationSec: number): Promise<Blob>;
  generateVoice(text: string): Promise<Blob>;
}

// MusicGenProvider | PiperProvider | ProceduralProvider (Web Audio)

ProceduralProvider uses Web Audio API oscillators + noise — zero API, always works offline. AI is optional enhancement, not a hard dependency.

Ethics & labeling

AI-generated audio is always labeled in the UI. Monastery-reviewed catalog stays separate from generated content. Users can practice with recorded dharma only — no AI required.