Getting Voice Onto a CPU (Part 2): More Attempts at Qwen3 TTS11 July 2026·1101 words·6 minsProjects Voice-Assistant Local-Ai Asr Tts Onnx Gguf Edge Cpu Voice-CloningONNX quantisation tuning produced great RTF numbers and unusable audio. A GGUF runtime with worse numbers won on the only test that mattered.
Getting Voice Onto a CPU: CPU-Only ASR and TTS for Fulloch26 June 2026·3492 words·17 minsProjects Voice-Assistant Local-Ai Asr Tts Onnx Edge CpuHow Fulloch’s CPU tier runs speech recognition and text-to-speech on a CPU-only PC
Watching a Voice Assistant Get Dumber: Qwen3.5 from 9B Down to 0.8B18 June 2026·2303 words·11 minsProjects Voice-Assistant Local-Ai Llm Quantization Self-Hosted BenchmarksSame conversation, same prompt, same tools, four model sizes. Watching Fulloch degrade from a capable agent into something that confuses the World Cup with Formula 1.
A Year Building a Fully Local Home Voice Assistant14 June 2026·3482 words·17 minsProjects Voice-Assistant Local-Ai Llm Home-Assistant Self-HostedA first year development diary of Fulloch, the Fully Local Home voice assistant.