arenasys / qDiffusion

Qt GUI for Stable diffusion
78 stars 5 forks source link

Qt GUI for Stable diffusion


Built from the ground up alongside sd-inference-server, the backend for this GUI. example *new* Discord: Arena Systems.

Getting started

Install

  1. Download this repo as a zip and extract it.
  2. Run qDiffusion.exe (or bash ./source/start.sh on Linux, sh ./source/start-mac.sh on Mac).
    • First time users will need to wait for Python and PyQt5 to be downloaded.
    • AMD Ubuntu users need to follow: Install ROCm.
  3. Select a mode. Remote, Nvidia and AMD are available.
    • Remote needs ~500MB of space, NVIDIA/AMD need ~5-10GB.
    • Choose Remote if you only want to generate using cloud/server instances.
    • For local generation choose NVIDIA or AMD, they also have the capabilities of Remote.
    • AMD on Windows uses DirectML so is much slower than on Linux.
  4. Press Install. Requirements will be downloaded.
    • Output is displayed on screen, fatal errors are written to crash.log.
  5. Done. NOTE: Update using File->Update or Settings->Program->Update.

Information is available on the Wiki.

Remote

Notebooks for running a remote instance are available: Colab, Kaggle, SageMaker

  1. Install qDiffusion, this runs locally on your machine and connects to the backend server.
    • If using Mobile then skip this step.
  2. Open the Colab notebook. Requires a Google account.
  3. Press the play button in the top left. Colab may take some time to configure a machine for you.
  4. Accept or reject the Google Drive permission popup.
    • Accepting will mean models are saved/loaded from qDiffusion/models on your drive.
    • Rejecting will mean models are local, you will need to download them again next time.
  5. Wait for the requirements to be downloaded and the server to start (scroll down).
  6. Click the DESKTOP link to start qDiffusion and/or connect.
    • Alternatively copy the Endpoint and Password to qDiffusion under Settings->Remote, press Connect.
  7. Done. See Downloads for how to get models onto the instance.
    • Remaking the instance is done via Runtime->Disconnect and delete runtime, then close the tab and start from Step 1.
    • HTTP 530 means the cloudflare tunnel is not working. Wait for an update, or check Here.
    • Runtime disconnects due to "disallowed code" can happen occasionally, often when merging. For now these don't appear to be targeted at qDiffusion specifically.

Mobile

qDiffusion Web is available for mobile users. Features are limited compared to the full GUI (txt2img only).

Overview