Plat'sPortfolio
I experiment with image generation, speech synthesis, and language models, and build tools.

Models & software projects
Trained models, personal projects, and more.
Pixel-space Flow Matching image generation
A Flow Matching model that generates illustrations directly in pixel space without latent compression. The 380M-parameter U-JiT architecture supports multiple resolutions around 512px.
Technical articles & blog posts
Articles from Zenn and my personal blog.
01
Zenn / SPEECH SYNTHESIS
02Learning a voice with Irodori-TTS using only embeddings lighter than an audio file
Zenn / IMAGE GENERATION
03Pretraining a pixel-space Flow Matching model for illustrations
Zenn / SOFTWARE
04Using Apple Foundation Models through an OpenAI-compatible API
Blog / OSINT / CTF
05SWIMMER OSINT CTF 2026 writeup — team narehate
Zenn / IMAGE GENERATION
06Training a pixel-space Flow Matching model on illustrated faces
Blog / BROWSER EXTENSION
07How YomiComiX translates manga inside the browser
Blog / WEB SECURITY
08Reporting an authentication vulnerability in netkeiba
Zenn / LANGUAGE MODELS
Building an LLM to translate Japanese into illustration prompts
ZENN + PLATFORM / BLOGOriginal articles are written in Japanese.
About me
I’m Plat. I like illustration and machine learning, and I build and train models for image generation, speech synthesis, and language modeling. I’m also interested in security and OSINT, and report vulnerabilities and take part in OSINT CTFs.
Areas of interest
- 01Image generation & Flow Matching
- 02Speech synthesis & multimodality
- 03Language models & prompt generation
- 04Software & web development
- 05OSINT CTFs & web security
Accounts
Links to SNS accounts and other platforms.