Decided on Parakeet ONNX Runtime. Works pretty great. Realtime voice chat possible now. UX lacking.

stt-parakeet/requirements-stt.txt

@@ -0,0 +1,54 @@
+# Parakeet ASR WebSocket Server - Strict Requirements
+# Python version: 3.11.14
+# pip version: 25.3
+#
+# Installation:
+#   python3.11 -m venv venv
+#   source venv/bin/activate
+#   pip install --upgrade pip==25.3
+#   pip install -r requirements-stt.txt
+#
+# System requirements:
+#   - CUDA 12.x compatible GPU (optional, for GPU acceleration)
+#   - Linux (tested on Arch Linux)
+#   - ~6GB VRAM for GPU inference
+#
+# Generated: 2026-01-18
+
+anyio==4.12.1
+certifi==2026.1.4
+cffi==2.0.0
+click==8.3.1
+coloredlogs==15.0.1
+filelock==3.20.3
+flatbuffers==25.12.19
+fsspec==2026.1.0
+h11==0.16.0
+hf-xet==1.2.0
+httpcore==1.0.9
+httpx==0.28.1
+huggingface_hub==1.3.2
+humanfriendly==10.0
+idna==3.11
+mpmath==1.3.0
+numpy==1.26.4
+nvidia-cublas-cu12==12.9.1.4
+nvidia-cuda-nvrtc-cu12==12.9.86
+nvidia-cuda-runtime-cu12==12.9.79
+nvidia-cudnn-cu12==9.18.0.77
+nvidia-cufft-cu12==11.4.1.4
+nvidia-nvjitlink-cu12==12.9.86
+onnx-asr==0.10.1
+onnxruntime-gpu==1.23.2
+packaging==25.0
+protobuf==6.33.4
+pycparser==2.23
+PyYAML==6.0.3
+shellingham==1.5.4
+sounddevice==0.5.3
+soundfile==0.13.1
+sympy==1.14.0
+tqdm==4.67.1
+typer-slim==0.21.1
+typing_extensions==4.15.0
+websockets==16.0