From 763eb448330417158027c4e7b487beabb2d30be2 Mon Sep 17 00:00:00 2001 From: haixuantao Date: Sat, 2 Aug 2025 15:13:13 +0800 Subject: [PATCH] Use quantized model instead of fp16 for faster response and lower memory footprint --- examples/openai-realtime/whisper-template-metal.yml | 1 + 1 file changed, 1 insertion(+) diff --git a/examples/openai-realtime/whisper-template-metal.yml b/examples/openai-realtime/whisper-template-metal.yml index 394836b9..e1ef95f9 100644 --- a/examples/openai-realtime/whisper-template-metal.yml +++ b/examples/openai-realtime/whisper-template-metal.yml @@ -43,6 +43,7 @@ nodes: - text env: MODEL_NAME_OR_PATH: Qwen/Qwen2.5-0.5B-Instruct-GGUF + MODEL_FILE_PATTERN: "*[qQ]6_[kK].[gG][gG][uU][fF]" - id: tts build: pip install -e ../../node-hub/dora-kokoro-tts