#AI

VRAM per quantization for the 20B Qwen-Image-Edit-2511 in one table: FP8 from 6GB, NF4 16–20GB, BF16 24GB+, GGUF Q4 CPU-runnable. Minimum RTX 3060 12GB and recommended RTX 4070 Ti 16GB builds, plus the 96-angle Multi-Angle LoRA.

Comparing how to give AI characters personality across the Gemini, Claude, and OpenAI APIs — covering System Prompts, Prefill, and structured output.

An overview of Vercel Labs' agent-browser and how it can be used from Claude Code as a lightweight alternative to Playwright MCP.

Emotion recognition used to mean fighting with old native libraries. Today there are cloud APIs and local libraries, but one major vendor has already left the field for ethical reasons.

How to implement browser-based voice input using the Web Speech API and MediaRecorder.

Aiming for a characterful AI with an avatar and voice chat, I started by comparing voice APIs.

A reference sheet of camera angles, shot sizes, and camera effects for AI image generation — prompts alongside output images.

Tried the '5 must-have Claude Code settings' going viral on X to escape the constant Ask prompts — here's where the gap between expectations and reality shows up.

Experimenting with Hyper3D Rodin's multi-image input to improve 3D generation quality — and how T-pose vs. standing pose changed the results.

Using Hyper3D Rodin to generate 3D models from text prompts and images — plus tracking down and fixing a bug encountered along the way.

How to have Claude AI operate Blender with MCP. Setup and basic usage.

Planning a Japanese voice chat system built on the voice-chat project, using Gemini 2.0 Flash for LLM and VOICEVOX for TTS.