Service

CodeRick dev studio

Images, 3D models and game assets on your own hardware: no clouds, no subscriptions, no queues at someone else's service. We install the studio on your server; the team works in a browser, scripts and AI agents use the API.

CodeRick dev studio is a ready-made AI studio that we install on your server. Artists, designers and developers work in it through a browser, while scripts and AI agents use the API. All data stays with you: not a single image, prompt or model leaves your network.

Type a few words and get a 2K image. Give it an image and get a textured 3D model. Give it a model and it splits itself into parts.

Who it is for

  • Game studios and indie teams — quick concepts, game items (weapons, armour, props), 3D blanks for prototypes and batch asset generation from templates.
  • Design bureaus and ad agencies — visuals, key frames, thumbnails and covers, edits from references.
  • YouTube creators and media — 1920×1080 thumbnails, Shorts, banners and avatars in one click.
  • Companies with strict privacy requirements — everything runs inside your perimeter, no external APIs.
  • Teams with AI agents — agents order images and models from the studio themselves through the API.

Images

  • 2K generation (about 4 megapixels) on Qwen-Image 2.1: photorealism, concept art, game assets.
  • Writes text right on the image well, Cyrillic included.
  • Smart prompt enhancer: write however it comes out, the network fills in the scene itself.
  • Up to 10 references: “dress the person from the first image in the jacket from the second”, “replace the background”, “draw it in this style”.
  • “Style only” reference: the style is carried over, the shape of the sample is not.
  • Background removal with one checkbox: a transparent PNG cropped to the object.
  • Ready-made formats: YouTube thumbnail, Shorts, banner, avatar — exact size and file weight.
  • Upscaling to 4K with detail refinement.

3D models

  • 3D from a single image or from a text description (TRELLIS.2).
  • PBR textures, normal map and ambient occlusion. You set the polygon count yourself — from light game models to detailed ones for rendering and printing.
  • Sharp texture: the source image is projected onto the model, so the pattern doesn’t “float”.
  • Back view: the network draws the back side of the object and checks the result, so the model has no “blind” back.
  • Matte, semi-matte or glossy materials — your choice.
  • Complex objects don’t bring the server down: the studio estimates the complexity and offers a choice of “more detail or faster”.
  • Export to GLB, OBJ (with textures), STL, PLY.

Splitting a model into parts

  • The model splits itself into parts (head, arms, weapon, armour pieces); you can say in text what to separate.
  • Every part can be downloaded on its own — for rigging, customisation and printing in pieces.
  • Works with your own models too: GLB, OBJ, FBX, STL and other formats.

Made for teamwork

  • Web interface in the browser — nothing to install on workstations. Light and dark theme, Russian and English.
  • Accounts and roles: every member has their own history, favourites and saved settings. The administrator sees everything.
  • Honest queue: you see what the GPU is doing right now, when your task starts and how long it will take.
  • Templates and settings profiles: good parameters are saved and repeated in one click.
  • “Reject” mark with a reason: failed results pile up as data for tuning generation to your tasks.
  • Tasks survive a service restart, results are never lost.

API for scripts and AI agents

  • Full HTTP API and a ready Python client: images, 3D, parts, history, export.
  • Batch generation: hundreds of assets from a list, with task labels, time estimates and automatic retry after a service restart.
  • Detailed documentation (in Russian) right inside the studio; the documentation version is tracked automatically.
  • Proven in practice: AI agents assembled sets of armour, weapons and vehicle parts for game projects through the API on their own.

Security and privacy

  • Everything runs on your server, no external services needed.
  • Password login, brute-force protection, login log, HTTPS for access from the internet.
  • Deleted images and models are wiped, not just deleted.
  • Users don’t see other people’s work unless they are given the rights.

What’s included

  • Installation and setup on your server (remotely).
  • All models, the web interface, API, Python client.
  • Documentation for users, administrators and developers.
  • Autostart and backups.
  • Technical support and new versions of the service for a year.

FAQ

Do I need the internet? Only for installation and updates. Generation runs inside your network.

Can I use the results in commercial projects? Every model comes with its own licence. There are two engines for splitting into parts: P3-SAM allows commercial use except in the EU, the UK and South Korea; PartField is non-commercial only. The studio shows the licence next to every result.

How many people can work at once? As many accounts as you like. Tasks line up in a shared GPU queue, and everyone sees when their task will start.

Is 24 GB of VRAM enough? Yes: on 24 GB the studio runs in economy mode and shares the memory between tasks by itself. Everything works, but 3D takes longer. The studio has been tested on both configurations from the table below.

Can I connect my own AI agents or scripts? Yes, through the API and the ready client.

How do I order? Press “Request a quote” and write to us — we’ll discuss your hardware and tasks.

System requirements

MinimumRecommended
Graphics cardNVIDIA with 24 GB VRAM (RTX 4090)96 GB (RTX PRO 6000 Blackwell)
RAM64 GB128 GB
DiskNVMe, 500 GB+ for the modelsNVMe 1–2 TB
OSWindows 11 or a Linux serverWindows 11 Pro or a Linux server