Auto Secure Login logoAuto Secure LoginPlatform Discord
Menu

AUTO SECURE LOGIN · HELP CENTER

ASL ImageLab setup and help

Create pictures with your own NVIDIA or AMD graphics card. Keep your library on your computer, save the settings behind each picture, and optionally make your workstation available to an authorized rBroker fleet.

Support: support@autosecurelogin.com. This guide is included with the app and works offline. External references and email require an internet connection.

This guide describes the release candidate currently being tested. The Store release is not yet available.

1. Set up your studio

  1. Open Setup. Choose a library folder on a local drive with enough free space. For an existing ImageLab installation, select the folder containing models, sd-cpp and out. Existing pictures and models are kept.
  2. Select AMD / Vulkan for a compatible AMD GPU, or NVIDIA / CUDA 12 for a compatible NVIDIA GPU. Install current graphics drivers from your hardware manufacturer. ImageLab does not replace drivers.
  3. Read the model licenses and accept their restrictions. Select Download missing files. Quality mode uses the SDXL base model and decoder, about 7.3 GB together. The optional Fast model adds about 6.9 GB. Engine downloads add roughly 30 MB for Vulkan or 915 MB for CUDA. Allow at least 25 GB free for the complete library and download files.
  4. Progress shows the current file and amount received. Cancel preserves the partial download for the next attempt. Every newly downloaded asset must match its pinned SHA-256 before installation.
  5. Choose Save & open studio. The local studio opens inside the app; no separate browser or Python installation is needed.

SDXL needs a compatible GPU with at least 12 GB of dedicated memory; 16 GB is recommended. Keep 12 GB of system RAM available for image work. Larger pictures, additional models and concurrent software can require more. Shared system memory does not replace dedicated GPU memory. ImageLab does not silently switch to CPU or a paid cloud service.

A selected library on an unplugged drive cannot be used until the drive returns. Changing folders does not move or delete your old library. Copy your own library before choosing another drive if you want to move it.

2. Create, cancel and refine a picture

Describe the subject, setting, composition and style in Prompt. For example: “A red toy sailboat on a still mountain lake, watercolor illustration, daylight.” Expand Negative prompt to list unwanted elements. Optional face-reference controls are in their own expandable section. In a smaller window, scroll the settings column; Generate remains visible at the bottom.

Click a gallery picture for its prompt, seed, model, dimensions and creation time. Use settings restores its controls. New seed makes another variation. Download saves a copy where you choose. The arrow keys move between pictures in the viewer.

Optional face-reference tools appear for separately prepared libraries. They require consent from the person and appropriately licensed models. The desktop installer does not download InsightFace’s noncommercial pretrained face weights. A photo reference is a creative aid, not verified identity or guaranteed likeness.

3. Understand what is saved

FolderContents
modelsModel weights and decoder.
sd-cppPinned CUDA or Vulkan rendering engine.
outYour pictures and adjacent JSON files containing their generation settings.
facesOptional local face-reference profiles, if installed and used.
.cachePartial downloads, temporary previews, render lock and local job diagnostics.

Application settings and desktop logs are separate from your library. Packaged installations use the app’s LocalState\ImageLab folder; unpackaged installations use %LOCALAPPDATA%\ASL\ImageLab. A small installation pointer under %USERPROFILE%\.asl\integrations\imagelab lets rBroker find this installation.

Use Setup → Troubleshooting → Open diagnostics folder to reach the correct logs. Uninstalling the app does not intentionally erase a library you chose outside its package. Back up pictures and recipes together. Never delete a lock or change library files while a render or download is using them.

4. Connect the existing rBroker

ImageLab is a complete local image application. rBroker is a separate resource coordination product. Its GPU lease reserves a memory budget for a named task; its heartbeat proves the task still owns that permission.

  1. Install and open the current ASL rBroker on this computer. Configure its GPU authority normally.
  2. In ImageLab Setup choose Find installed rBroker. Confirm the detected local runtime and GPU configuration. For an existing technical installation, expand Manual connection settings and browse to that runtime and its configuration file. Never select or paste the token file.
  3. Select Use this computer’s existing rBroker GPU leases, then save and open the studio.
  4. The header identifies the connection. Each local ImageLab render now requests its own 10 GiB GPU lease with a readable description, keeps the heartbeat active, and releases after stopping. Permission rejection or heartbeat loss stops the owned work; ImageLab does not bypass it.

Connecting ImageLab to the local GPU broker does not enroll a computer in a fleet and does not authorize remote callers. Those permissions belong to rBroker. Do not start a second GPU authority on a computer that already has one.

5. Let permitted computers request images

  1. On the rendering computer, install and configure ImageLab first. Verify a local picture works.
  2. In rBroker’s picture setup, choose Connect installed ASL ImageLab. rBroker validates the installed renderer, library and backend; it does not download models for you.
  3. Enable pictures in the local worker configuration. Update settings only while its owned executor and private worker are stopped or drained.
  4. Join your configured fleet server, enable image work in this computer’s local consent, and grant a named dispatcher permission to send image requests. Keep permissions bounded by the computer’s actual capabilities.
  5. The sender supplies a meaningful project, task, purpose and expected result. Both computers need a network route to the configured fleet service; they do not need to share a LAN.
  6. The receiver validates the request, reserves 12 GiB system RAM and a 10 GiB GPU lease for the ImageLab profile, renders one to four pictures, stops, and returns PNGs and a recipe through the encrypted result channel. The existing offer uses larger authorization ceilings; actual usage is separately verified.

Keep the owning rBroker receiver running until its accepted jobs settle and results transfer. Closing the ImageLab desktop window does not stop a separate fleet worker owned by rBroker. Blocking new fleet consent does not erase accepted work or bypass orderly cleanup. An offline sender may delay file-based jobs; requests with inline text already present can proceed if other admission requirements are met.

Multiple computers do not become a single larger graphics card. A model must fit the selected rendering computer. AI Hub can orchestrate permitted tasks, rBroker owns resources and transfer, and ImageLab performs image generation. Each remains separately usable.

6. Privacy, content and third-party licenses

Local prompts, images and optional face references are processed on your computer. Model downloads contact GitHub and Hugging Face. Enabling a fleet sends the selected task metadata and requested results through the configured rBroker workflow. You choose the fleet server and dispatcher permissions.

Do not generate illegal material, nonconsensual sexual images, harmful impersonations or material you have no right to use. Do not use another person’s face without permission. ImageLab’s Vulkan/CUDA engine does not include a generated-pixel safety classifier. Fleet requests use rBroker’s prompt checks, which are not a guarantee that every generated result is suitable. Review outputs before sharing them.

SDXL base and Lightning are subject to their model licenses and use restrictions. A purchase of ImageLab does not grant ownership of third-party intellectual property or guarantee rights in every output. InsightFace pretrained weights have separate noncommercial restrictions; they are not included in the desktop download. See the licenses from SDXL, SDXL-Lightning, and stable-diffusion.cpp.

7. Fix a problem using its specifics

What you seeWhat it means and what to do
Missing model or engineThe selected library lacks the named file. Select the correct existing library or download missing assets in Setup. Quality and Fast use different models.
Library unavailableThe configured drive or folder cannot be read. Reconnect it or select an available folder. Existing pictures have not been moved.
Download checksum failureThe downloaded bytes do not match the pinned file. They are not installed. Check the network/proxy and remove only the named failed partial file before retrying.
No CUDA/Vulkan device, engine exitThe selected backend cannot run with this GPU/driver or required runtime files are missing. Confirm the correct backend and manufacturer driver. Consult the timestamped error; no CPU fallback occurred.
Waiting for the GPUAnother ImageLab process owns the library’s render lock, or the broker has not admitted this work. Let that work finish, or cancel your own request. Do not delete lock files or terminate unknown processes.
GPU capacity or RAM refusalThe requested task cannot safely fit alongside existing work. Close optional applications or wait. A large Windows “shared GPU memory” figure does not increase the dedicated GPU budget.
Authentication rejection or heartbeat failureThe existing broker did not confirm ownership. ImageLab stops the owned work. Check the broker’s live status, configured authority and credential access in the owning account. Do not borrow another task’s lease.
Fleet sees the computer but cannot renderMonitoring is not execution permission. Check local ImageLab setup, enabled image handler, unexpired node consent and the sender’s dispatcher grant.
Picture rendered but not deliveredThe result remains on the receiver while transfer or receipt verification is unfinished. Resume the original saved job with both endpoints available; do not send a duplicate to conceal the failure.
Studio cannot openRead the timestamped startup error. Repair the Microsoft Edge WebView2 runtime if it is absent, or reconnect the selected library/broker. Open Setup rather than launching a second hidden backend.

8. Get useful support

Email support@autosecurelogin.com. Include ImageLab and rBroker versions, Windows version, GPU model, selected backend, the exact failure time with timezone, what you clicked, and whether the problem occurs locally or only through a fleet. Say whether work is still running.

Open diagnostics folder shows local desktop logs. Review excerpts before sending them: prompts, file paths or names may be private. Do not send credentials, fleet keys, signing keys, whole configuration/state folders or private face photographs unless a separately agreed support process requires them.

Technical references: Microsoft WebView2, stable-diffusion.cpp. Support cannot guarantee a model’s exact likeness, text accuracy or reproducibility across different hardware.

Help / Ayuda