danielrosehill

FFmpeg GPU-Accelerated Video Compression (VA-API / AMD)

Compress video using hardware-accelerated H.264 encoding on AMD GPUs via VA-API.

Tested on: AMD Radeon RX 7700 XT (Navi 32, gfx1101) / Ubuntu 25.04

Basic Compression Command

ffmpeg -vaapi_device /dev/dri/renderD128 \

danielrosehill

Fixing OpenCL/GPU Acceleration for AMD Radeon on Ubuntu

Problem

Video editors (Kdenlive, DaVinci Resolve, etc.) had broken GPU acceleration on Ubuntu 25.04 with AMD Radeon RX 7700 XT.

Running clinfo showed:

Number of platforms: 0

danielrosehill

Kdenlive 25.08.3 VAAPI AMD Crash Report

Environment

• OS: Ubuntu 25.04
• Desktop: KDE Plasma (Wayland)
• Kdenlive: 25.08.3 (Flatpak from Flathub, stable branch)
• GPU: AMD Radeon RX 7700 XT (Navi 32, gfx1101)
• Kernel: 6.14.0-15-generic
• Mesa: 25.2.3-1ubuntu1

danielrosehill

Adding MCP Router to Claude Code CLI

MCP Router allows you to manage MCP servers across multiple AI applications from a single interface.

Installation Command

claude mcp add mcp-router --scope user -e "MCPR_TOKEN=your_token_here" -- npx -y @mcp_router/cli@latest connect

danielrosehill

Snapcast Multi-Room Audio Setup with Music Assistant and Home Assistant

This guide documents setting up Snapcast for multi-room audio with Music Assistant integration on a home network.

Architecture

• Snapserver: Docker container on Ubuntu VM (10.0.0.4)
• Snapclients: Raspberry Pi devices with USB speakers
• Control: Home Assistant (10.0.0.3) with Music Assistant integration

danielrosehill

F13 Key Not Recognized on KDE Plasma (Wayland) - Troubleshooting Notes

Problem

F13 key (sent via input-remapper from a macropad) was not being recognized by KDE for keyboard shortcuts, while F14-F18 worked fine.

Environment

• OS: Ubuntu 25.04
• Desktop: KDE Plasma on Wayland

danielrosehill

Extended Function Keys (F13-F35) on Linux

A reference guide for mapping keys beyond F12 on Linux systems.

Summary

Layer	F-Key Range	Notes
Linux Kernel	F1-F24	Native evdev keycodes
XKB (X11/Wayland)	F1-F24	Mapped via evdev keycodes

danielrosehill

CH57x 6-Button Macropad Setup Guide (Linux)

Product Details

Field	Value
Product Name	Mechanical Hotswap Macropad Custom Keyboard USB Bluetooth RGB Photoshop Gaming Programming Macro Knob Keypad
Seller	Romoral

danielrosehill

Whisper GGML Model Sizes for FUTO Voice Input

Reference for choosing Whisper model sizes when using custom fine-tuned models with FUTO Voice Input on Android.

Model Size Comparison (GGML f16)

Model	Parameters	Approx GGML Size
Tiny	39M	~75 MB
Base	74M	~142 MB

danielrosehill

Choosing Epochs for ASR Fine-Tuning

A practical guide to selecting the right number of training epochs when fine-tuning speech recognition models like Whisper, Wav2Vec2, HuBERT, and similar ASR architectures.

What's Different About ASR Fine-Tuning?

ASR models have unique characteristics that affect epoch selection:

1. Pre-trained on massive audio data - Whisper was trained on 680,000 hours of audio
2. Audio samples vary in length - 2 seconds to 30 seconds per sample