39 lines
889 B
Plaintext
39 lines
889 B
Plaintext
LOCAL_AI_VENV_PATH=.local-ai/.venv
|
|
# Windows면 보통 python 또는 py -3
|
|
LOCAL_AI_PYTHON=python
|
|
|
|
# Windows: ffmpeg dshow 장치 이름
|
|
# Linux: pactl list sources short 에서 monitor/source 이름
|
|
AUDIO_SOURCE=
|
|
DOCKER_BIN=
|
|
|
|
DEBUG=false
|
|
TTS_ENABLED=true
|
|
TTS_IMAGE=realtime-voice-bot-melotts:v0.1.2
|
|
TTS_LANGUAGE=KR
|
|
TTS_SPEAKER=KR
|
|
TTS_DEVICE=cpu
|
|
TTS_SPEED=1
|
|
TTS_CACHE_DIR=.local-ai/tts-cache
|
|
TTS_OUTPUT_DIR=.local-ai/tts-output
|
|
OLLAMA_BASE_URL=http://127.0.0.1:11434
|
|
OLLAMA_MODEL=qwen3:8b
|
|
OLLAMA_KEEP_ALIVE=5m
|
|
MAX_CONVERSATION_TURNS=6
|
|
WHISPER_MODEL=large-v3-turbo
|
|
WHISPER_LANGUAGE=ko
|
|
WHISPER_DEVICE=auto
|
|
WHISPER_COMPUTE_TYPE=auto
|
|
WHISPER_BEAM_SIZE=2
|
|
|
|
SEGMENT_START_THRESHOLD=900
|
|
SEGMENT_CONTINUE_THRESHOLD=450
|
|
SEGMENT_START_FRAMES=2
|
|
SEGMENT_END_FRAMES=24
|
|
SEGMENT_PREROLL_SAMPLES=3200
|
|
SEGMENT_MIN_SPEECH_SAMPLES=7200
|
|
SEGMENT_MAX_SPEECH_SAMPLES=160000
|
|
|
|
DEBUG_TRANSCRIPTS=true
|
|
LOG_LEVEL=info
|