Home

Independent AI and IT Studio

Model research with practical engineering discipline.

PrySai Lab focuses on AI image model R&D around Illustrious, NoobAI, Pony, FLUX, and Z-Image format workflows, while building applied work around LLM content systems and memory-oriented AI tooling. We prioritize transparent scope, reproducible process, and steady iteration over inflated claims.


Measured usage from the last year.

Across public and online generation channels, recent usage has accumulated into a meaningful operating baseline. These figures describe activity volume and delivery scale, not quality claims or market ranking.

Based on the last 12 months of service activity across public and hosted channels.

Online generations started

0M+

In the last year, online image generation services have been started more than 7 million times across supported workflows and service channels.

AI artworks generated

0M+

Generated output has exceeded 10 million AI artworks, reflecting sustained usage rather than one-time promotional volume.

Users served

0K+

Services and model access have reached more than 50 thousand users through open distribution and online generation products.

Current public channels and service availability.

At present, PrySai Lab mainly provides open-source distribution on CivitAi and online generation plus customized subscription services on SeaArt. These channels serve different user needs and access patterns.

CivitAi

PrySai Lab currently provides open-source model access and public distribution through CivitAi for users who prefer direct discovery, download, and community-based iteration.

Open-source profile

SeaArt

SeaArt is currently used to provide online generation access and customized subscription-oriented services, offering a more direct hosted experience for end users.

Service profile

What we are building now

Current work stays grounded in measurable tasks: model behavior tuning, workflow tooling, and content architecture for language systems. We document constraints and trade-offs so progress remains honest and useful.

Image Model Pipeline

Training and evaluation workflows for diffusion-style and related creative model formats, with attention to consistency, controllability, and deployment fit.

LLM Content Systems

Applied language model structures for studio documentation, assistant-like workflows, and practical prompt and process standardization.

AI Memory Layer

Early-stage memory modules for context persistence and retrieval quality, aimed at improving continuity in agent and assistant scenarios.

Illustrious NoobAI Pony FLUX Z-Image LLM

Objective by default, ambitious by method.

We do not present prototypes as finished products. Public-facing statements are kept factual, and outputs are framed by current maturity and known limitations.

Scope Control

Clear boundaries for each cycle: what is being tested, what is shipped, and what remains experimental.

Documentation First

Design and model decisions are recorded to improve repeatability and reduce handoff risk.

Infrastructure Awareness

Web delivery, caching, and runtime constraints are treated as part of product quality, not afterthoughts.