Or jump from this list on a smaller screen.

This is a page from the Audivra documentation. Developer exportsindex: http://127.0.0.1:3000/docs/llms.txt, full corpus: http://127.0.0.1:3000/docs/llms-full.txt, OpenAPI: http://127.0.0.1:3000/docs/openapi.json, Postman: http://127.0.0.1:3000/docs/postman.json. Press ⌘K to search docs.

Local worker

Self-hosted open-source stack — Ollama, Kokoro-82M, MuseTalk on RTX 3060/4090 or CPU.

Short prompts route to local Ollama (LLM), Kokoro/Piper (TTS), and MuseTalk/Wav2Lip (lip-sync) before offloading to GPU cluster or external APIs. Every execution is logged in PostgreSQL with atomic credit deduction.

Stack

  • Fast LLMOllama — qwen2.5:3b / llama3.2:3b (~2.5 GB VRAM/RAM)
  • TTSKokoro-82M or Piper TTS (~500 MB RAM / CPU-friendly)
  • Lip-sync videoMuseTalk or Wav2Lip (~4 GB VRAM)

Setup

Terminal
# LLM (fast path)
ollama pull llama3.2:3b

# Inference (TTS + lip-sync)
pip install -r inference/requirements.txt
npm run inference          # :8090

# Optional unified gateway
npm run local-worker       # :11435

Environment

.env
OLLAMA_URL=http://127.0.0.1:11434
LOCAL_WORKER_GATEWAY_URL=http://127.0.0.1:11435
INFERENCE_URL=http://127.0.0.1:8090
FAST_PATH_CHAR_LIMIT=500
LOCAL_VRAM_THRESHOLD_GB=6

Routing logic: Processing nodes · API: POST /api/v1/generate

Step-by-step guide: Self-host on an RTX 4090