shinji shimizu

i am a ai web full stack dev come to my page kotonia.ai

japan Joined on May 22, 2026 https://kotonia.ai/

Pronouns

MAN

shinji shimizu

May 23

HiDream Raw Output Failed Tried Dev-2604 VRAM Math Killed It Won with a Prompt Enhancer Instead

#hidream #diffusion #promptengineering #gemini

10 min read

shinji shimizu

May 22

Building a Sarcastic AI English Tutor with Persona-as-Code and Gemini Audio Input for Pronunciation Correction

#ai #webdev #typescript #rust

7 min read

shinji shimizu

May 22

Five Years Later, I Finally Have 96GB VRAM — What It Actually Unlocks for Agent Loops

#gpu #ai #machinelearning #python

8 min read

shinji shimizu

May 22

Turning a 1-Line Idea Into a 40-Second Short with a 10-Beat Local Video Pipeline

#python #ai #machinelearning #gpu

7 min read

shinji shimizu

May 22

Running LTX-2.3 Alongside TTS on a Single 96GB GPU with a Cold-Start Architecture

#gpu #python #machinelearning #ai

5 min read

shinji shimizu

May 22

Cutting LTX-2 22B Peak VRAM by 40% with fp8_cast — and Why optimum-quanto Was a Trap

#ai #machinelearning #gpu #python

7 min read

shinji shimizu

May 22

HiDream Skeleton Mode: Prompt Beats OpenPose Ref — 8 Patterns Benchmarked

#ai #python #machinelearning #gpu

11 min read

shinji shimizu

May 22

Replicating a Language-Learning Comedy Short with Claude Code — Gemini as a Multimodal Sub-Agent

#ai #python #machinelearning #productivity

8 min read

shinji shimizu

May 22

HiDream-O1-Image 3–8x Faster: Benchmarking Steps, CFG, and Resolution

#ai #machinelearning #gpu #python

5 min read

Forem

shinji shimizu

Badges

Writing Debut

Skills/Languages

Currently learning

Currently hacking on

HiDream Raw Output Failed Tried Dev-2604 VRAM Math Killed It Won with a Prompt Enhancer Instead

Building a Sarcastic AI English Tutor with Persona-as-Code and Gemini Audio Input for Pronunciation Correction

Five Years Later, I Finally Have 96GB VRAM — What It Actually Unlocks for Agent Loops

Turning a 1-Line Idea Into a 40-Second Short with a 10-Beat Local Video Pipeline

Running LTX-2.3 Alongside TTS on a Single 96GB GPU with a Cold-Start Architecture

Cutting LTX-2 22B Peak VRAM by 40% with fp8_cast — and Why optimum-quanto Was a Trap

HiDream Skeleton Mode: Prompt Beats OpenPose Ref — 8 Patterns Benchmarked

Replicating a Language-Learning Comedy Short with Claude Code — Gemini as a Multimodal Sub-Agent

HiDream-O1-Image 3–8x Faster: Benchmarking Steps, CFG, and Resolution