Stable Diffusion Worker Retirement

The standalone Stable Diffusion Worker (Automatic1111) is retired as of October 27, 2026. Use Sogni Comfy Worker for Stable Diffusion 1.5 and SDXL workloads.

Stable Diffusion models remain available. This retirement concerns the separate worker application and Docker images, including its no-models variant.

#Move to Comfy Worker

1. Check the Comfy hardware requirements. On x86-64, the current installer requires an NVIDIA Ampere-or-newer GPU and driver R580 or newer.
2. In the Workers Dashboard, choose Stop after current job. Wait for Safe to shut down, then stop the old worker through its installer or hosting provider. Disable its automatic restart before reusing its NFT.
3. Keep the old installation, configuration, and model folders as a backup. Download the Windows installer or Linux installer into a separate persistent folder.
4. Follow the local installation guide and choose Comfy Worker (1). Reuse your Sogni account API key and NFT token ID only after the old worker has stopped.
5. At the model prompt, choose Current models plus the classic Stable Diffusion 1.5 / SDXL library, or Only the classic Stable Diffusion 1.5 / SDXL library. Comfy downloads its own required assets. Allow enough disk space for both installations during migration.
6. Confirm the same NFT reconnects as a Comfy worker in the Dashboard. Check loaded models, health, and completed jobs before removing any old installation data.

An installer update alone does not convert an existing SD worker. Existing settings and model files are preserved; the old Automatic1111 cache is not automatically imported into Comfy.

#Hosted and multi-GPU workers

Use the remote deployment guide and the appropriate Comfy image, storage mounts, and health probes. Set the existing Comfy setting MODEL_COLLECTIONS=sd for the classic library or MODEL_COLLECTIONS=modern,sd to include current models. SD_COLLECTION_MAX_GB defaults to a 150 GB checkpoint budget; shared assets need additional space. See model collections.

For multiple GPUs, configure the new installation with one unique NFT per GPU and migrate each identity only after its old worker has stopped. Rerunning the installer against an existing generated multi-GPU configuration preserves that configuration; it does not convert its worker type.

Do not deploy the retired sogni/sogni-stable-diffusion-worker, sogni/sogni-stable-diffusion-worker-no-models, or Flux worker images. Use a current Comfy recipe instead.

For migration help, email [email protected] or join the Sogni Discord.