# FLUX.1 for AUTOMATIC1111 Users: A Forge Migration Guide > Published 2026-04-07, updated 2026-09-05 · https://www.promptzone.com/mariam_kobayashi/flux-boosts-automatic1111-for-faster-ai-images-6b4 --- title: "FLUX.1 for AUTOMATIC1111 Users: A Forge Migration Guide" published: true description: "FLUX.1 for AUTOMATIC1111 users: install Forge, select the right model components, and rebuild a working generation setup with verified guidance." tags: ai,stablediffusion,imagegeneration,flux --- FLUX.1 is Black Forest Labs’ image-model family, and Forge is a WebUI-derived application with explicit support for loading its downloadable models. AUTOMATIC1111’s upstream README documents Stable Diffusion features but does not document native FLUX loading; the verified WebUI-style route described here is Forge. Obtain Forge from its maintainer’s repository and select a matching FLUX model package. [BFL announcement][launch], [AUTOMATIC1111 README][a1111], [Forge README][forge] This guide focuses on making that transition understandable for an existing AUTOMATIC1111 user. The first milestone is a working FLUX generation with known components, followed by deliberate reintroduction of the workflow features you need. ## What are the key facts about FLUX and AUTOMATIC1111? | Field | Verified detail | | --- | --- | | Developer | Black Forest Labs develops FLUX.1; lllyasviel’s project maintains Forge. [BFL][launch], [Forge][forge] | | Released | FLUX.1 launched August 1, 2024; Forge’s maintainer documented FLUX support in August 2024. [BFL][launch], [Forge tutorial][setup] | | Type | Text-to-image model family used through a local WebUI application. [Dev card][dev], [Forge][forge] | | Size or parameters | FLUX.1-dev’s transformer has 12 billion parameters. [Model card][dev] | | License and access | Model terms depend on the checkpoint: dev uses its Non-Commercial License; schnell declares Apache-2.0. Forge software is AGPL-3.0. [Dev][dev], [Schnell][schnell], [Forge][forge] | | Where it runs | Locally through Forge’s documented FLUX loading paths; hosted FLUX services are separate. [Forge][forge], [BFL][launch] | ## What does Forge offer AUTOMATIC1111 users running FLUX? Forge retains a WebUI-style application while exposing controls needed for FLUX loading. The maintainer’s separate-component tutorial explains changes to model, encoder, and VAE selection. That is useful for users who want familiar generation controls while making the model dependencies explicit. [Forge component guide][components] The project also documents loading packaged bitsandbytes NF4 checkpoints and separate full or GGUF model files. These are alternative loading arrangements, and each needs its matching instructions. Begin with one documented arrangement rather than mixing instructions from several formats. [Forge README][forge] For FLUX.1-dev, Forge distinguishes standard CFG from distilled guidance. Its maintainer gives a sample configuration and a sanity-check image so users can establish a common baseline. Follow that baseline before judging how a migrated prompt behaves. [Forge FLUX tutorial][setup] AUTOMATIC1111 itself documents generation metadata and tools for recovering settings from saved images. Before trying another application, use those features to record the original prompt and configuration you want to compare. This creates a clear starting point for the transition. [AUTOMATIC1111 README][a1111] ## What should you check before migrating to FLUX in Forge? Forge is a separate implementation. Its README says it is based on Stable Diffusion WebUI, and its maintainer describes additional component-selection controls for newer models. Similar-looking screens should not be taken as proof that the upstream AUTOMATIC1111 application supports the same architectures. [Forge README][forge], [Component guide][components] Compatibility also needs to be checked at the feature level. Forge publishes extension guidance and a status table with dated checks; that is not a blanket guarantee that every extension will work with every FLUX checkpoint. Test the specific feature you need after basic generation succeeds. [Forge README][forge] FLUX.1-dev can still produce images that miss prompt requirements. Its model card also notes sensitivity to prompting style. A changed output after migration may reflect the new model or settings, so compare behavior carefully rather than assuming a loading failure. [Model card][dev] Memory use and speed require measurement on the actual setup. Forge’s FLUX tutorial explains the tradeoffs between GPU weight allocation and offloading, and cautions that allocating too much can harm performance. No universal per-image timing follows from installing the interface. [Forge tutorial][setup] ## How do you migrate from AUTOMATIC1111 to FLUX in Forge? First, save a few representative AUTOMATIC1111 outputs with their generation settings. Choose examples that exercise the features you actually use, such as a simple portrait, an object composition, and a prompt with requested lettering. This is a suggested comparison set, not a benchmark corpus. Install Forge in its own directory using the official README. The maintainer offers a packaged Windows installation and a Git-based route for users who already have the documented Git and Python prerequisites. For the latter, the README gives this installation pattern: [Forge README][forge] ```bat git clone https://github.com/lllyasviel/stable-diffusion-webui-forge.git cd stable-diffusion-webui-forge webui-user.bat ``` The clone and launcher prepare the application; they do not select a FLUX checkpoint. Follow either the packaged NF4 tutorial or the separate-component tutorial next. Keep the two routes distinct while getting the first image to work. [Forge tutorials][setup], [Component guide][components] For the packaged route, the maintainer’s linked NF4 repository recommends `flux1-dev-bnb-nf4-v2.safetensors`. It contains the quantized model with accompanying encoder and VAE components. Select it through Forge’s checkpoint loader and initially leave loading precision on `Auto`. [NF4 card][nf4], [Forge tutorial][setup] For separate files, Forge documents the base model under `models/Stable-diffusion`, the VAE under `models/VAE`, and CLIP-L and T5 under `models/text_encoder`. Select the corresponding components in the UI. These paths come from Forge’s own tutorial and should not be replaced with directories from a different interface. [Component guide][components] For a FLUX.1-dev baseline, use the maintainer’s documented Euler and Simple sampling configuration with CFG at one and distilled guidance at 3.5. The tutorial explains that the negative-prompt field is inactive at that CFG setting. Treat that behavior as part of the selected pipeline. [Forge tutorial][setup] Generate a simple image before enabling any additional adapters or extensions. Then rebuild one required feature at a time, saving a result and its settings after each successful change. If something fails, return to the most recent working configuration and inspect the application’s error message. When comparing against AUTOMATIC1111, evaluate the deliverable rather than expecting matching pixels. Record subject placement, prompt adherence, troublesome details, and workflow effort. The purpose is to learn which application-model combination supports your work with the fewest unresolved problems. ## How do AUTOMATIC1111, Forge, and ComfyUI compare? | Option | What its primary documentation establishes | | --- | --- | | AUTOMATIC1111 upstream | Stable Diffusion generation and related WebUI features; no native FLUX loading procedure in the cited README. [README][a1111] | | Forge | Explicit FLUX loading tutorials, including packaged and separate-component routes. [README][forge] | | ComfyUI | Official FLUX workflows with visible model, encoder, and VAE nodes. [Tutorial][comfy] | For the graph-based alternative, read the [ComfyUI complete guide](/tomas_novak/comfyui-2026-the-complete-guide-to-power-user-ai-image-generation-1g17). For a general installation walkthrough, use the sibling [FLUX Forge tutorial](/lucia_arellano/flux-forge-ai-image-tool-tutorial-2k9e); keep this migration checklist focused on the prompts, settings, and extensions you are carrying over. ## What else should AUTOMATIC1111 users know about FLUX? ### How should AUTOMATIC1111 users access FLUX locally? Forge provides documented FLUX checkpoint and component-loading procedures for a WebUI-style workflow. The upstream AUTOMATIC1111 README does not provide a native FLUX loading procedure. [Forge README][forge], [AUTOMATIC1111 README][a1111] ### Is Forge the same project as AUTOMATIC1111? Forge builds on Stable Diffusion WebUI but maintains its own repository and model-loading changes. Configure Forge's FLUX features using the Forge documentation. [Forge README][forge] ### Can I reuse all my AUTOMATIC1111 extensions in Forge? Forge publishes extension guidance and a status table with dated checks. Check each extension you need and test it with your selected FLUX model after basic generation works. [Forge README][forge] ### Why is the negative prompt inactive for FLUX.1-dev in Forge? Forge's FLUX-dev tutorial recommends CFG at one, which greys out the negative-prompt field. Its baseline uses the separate distilled-guidance control at 3.5. [Forge tutorial][setup] ## Sources - [BFL’s original FLUX announcement][launch] - [AUTOMATIC1111 upstream README][a1111] - [Forge README][forge] - [Forge FLUX and NF4 tutorial][setup] - [Forge component-loading tutorial][components] - [FLUX.1-dev model card][dev] - [FLUX.1-schnell model card][schnell] - [lllyasviel NF4 checkpoint card][nf4] - [ComfyUI FLUX workflow][comfy] [launch]: https://bfl.ai/blog/24-08-01-bfl [a1111]: https://github.com/AUTOMATIC1111/stable-diffusion-webui [forge]: https://github.com/lllyasviel/stable-diffusion-webui-forge [setup]: https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/981 [components]: https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050 [dev]: https://huggingface.co/black-forest-labs/FLUX.1-dev [schnell]: https://huggingface.co/black-forest-labs/FLUX.1-schnell [nf4]: https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4 [comfy]: https://docs.comfy.org/tutorials/flux/flux-1-text-to-image ## Related guides on PromptZone - [Realistic Photos with FLUX](/stabletom/realistic-photos-with-flux-57aa) - [Best SDXL Models in 2026](/tara_suzuki/best-sdxl-models-in-2026-realistic-anime-and-all-purpose-checkpoints-116)