Install AngeVoice on fnOS

The important fnOS choices are the runtime profile and persistent directories, not a large number of tuning variables.

Before installation

• Confirm the NAS architecture and available acceleration. Choose CPU when no supported NVIDIA GPU is available.
• Reserve persistent storage for models and generated files.
• Enable FFmpeg only when MP3, OGG or M4A output is needed.

Installation flow

1. Install the AngeVoice FPK from the fnOS app flow or local package installer.
2. Choose CPU, standard GPU or Legacy GPU. When unsure, start with CPU.
3. Map models, configuration, credentials, prompts and outputs to persistent storage.
4. Start the app and wait for health to become ok or idle.
5. Change the default administrator credentials and save the API key.

Recommended runtime defaults

KOKORO_PROCESS_ISOLATION_ENABLED=true
MOSS_PROCESS_ISOLATION_ENABLED=true
ZIPVOICE_PROCESS_ISOLATION_ENABLED=true
ANGEVOICE_STARTUP_PRELOAD_ENABLED=false
ANGEVOICE_RESTART_AFTER_IDLE_UNLOAD=false

On-demand loading is a practical NAS default. Enable startup preload only when first-request latency matters more than memory use.

Audio conversion

ANGEVOICE_FFMPEG_ENABLED=true
ANGEVOICE_FFMPEG_BINARY=ffmpeg
ANGEVOICE_FFMPEG_TIMEOUT_SECONDS=30

Keep FFmpeg disabled when WAV or PCM is sufficient. For conversion errors, verify that the app can run ffmpeg -version.

Upgrade checklist

1. Back up credentials/, config/, prompts/ and required output files.
2. Record the active runtime profile and port.
3. After upgrading, check /health, /v1/models and one real synthesis request.
4. Confirm that the administrator account, API key and Voice Profiles still exist.