pepperplus/pepperplus-cb
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

feat: prepend more silence before speech audio for better transcription beginnings

Browse Source 
ref: N25B-429
This commit is contained in:
Twirre Meulenbelt
2026-01-08 10:49:13 +01:00
parent 5a61225c6f
commit 45719c580b
3 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.env.example
View File

@@ -10,7 +10,7 @@ LLM_SETTINGS__LOCAL_LLM_URL="http://localhost:1234/v1/chat/completions"
LLM_SETTINGS__LOCAL_LLM_MODEL="gpt-oss"
# Number of non-speech chunks to wait before speech ended. A chunk is approximately 31 ms. Increasing this number allows longer pauses in speech, but also increases response time.
BEHAVIOUR_SETTINGS__VAD_NON_SPEECH_PATIENCE_CHUNKS=3
BEHAVIOUR_SETTINGS__VAD_NON_SPEECH_PATIENCE_CHUNKS=15
# Timeout in milliseconds for socket polling. Increase this number if network latency/jitter is high, often the case when using Wi-Fi. Perhaps 500 ms. A symptom of this issue is transcriptions getting cut off.
BEHAVIOUR_SETTINGS__SOCKET_POLLER_TIMEOUT_MS=100