diff --git a/.dockerignore b/.dockerignore
new file mode 100644
index 000000000..827e02326
--- /dev/null
+++ b/.dockerignore
@@ -0,0 +1,8 @@
+.*
+!.git
+__pycache__/
+*.py[cod]
+input
+models
+notebooks
+output
diff --git a/.github/workflows/docker.yml b/.github/workflows/docker.yml
new file mode 100644
index 000000000..541a7aac9
--- /dev/null
+++ b/.github/workflows/docker.yml
@@ -0,0 +1,101 @@
+name: Build and publish Docker images
+
+on:
+  push:
+    branches:
+      - master
+  pull_request:
+    branches:
+      - master
+  release:
+    types: [published]
+
+jobs:
+  docker:
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - id: cu118
+            name: CUDA 11.8
+            pytorch_install_args: "--index-url https://download.pytorch.org/whl/cu118"
+          - id: cu121
+            name: CUDA 12.1
+            pytorch_install_args: "--index-url https://download.pytorch.org/whl/cu121"
+          - id: cu124
+            name: CUDA 12.4
+            pytorch_install_args: "--index-url https://download.pytorch.org/whl/cu124"
+          - id: rocm6.2
+            name: ROCm 6.2
+            pytorch_install_args: "--index-url https://download.pytorch.org/whl/rocm6.2"
+          - id: cpu
+            name: CPU only
+            pytorch_install_args: "--index-url https://download.pytorch.org/whl/cpu"
+            extra_args: --cpu
+
+
+    name: ${{ matrix.name }}
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Check which repositories to use
+        id: repositories
+        run: |
+          echo "GHCR_IMAGE_NAME=ghcr.io/${GITHUB_REPOSITORY_OWNER}/comfyui" >> "$GITHUB_ENV"
+          if [[ -n "${DOCKERHUB_USERNAME}" ]]; then
+            echo "DOCKERHUB_IMAGE_NAME=${DOCKERHUB_USERNAME}/comfyui" >> "$GITHUB_ENV"
+          else
+            echo "DOCKERHUB_IMAGE_NAME=" >> "$GITHUB_ENV"
+            echo "No Docker Hub username set, only deploying to GitHub Container Repository"
+          fi
+        env:
+            DOCKERHUB_USERNAME: ${{ secrets.DOCKERHUB_USERNAME }}
+            GITHUB_REPOSITORY_OWNER: ${{ github.repository_owner }}
+      - name: Docker meta
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          # list of Docker images to use as base name for tags
+          images: |
+            ${{ env.DOCKERHUB_IMAGE_NAME }}
+            ${{ env.GHCR_IMAGE_NAME }}
+          flavor: |
+            suffix=-${{ matrix.id }},onlatest=true
+          # generate Docker tags based on the following events/attributes
+          tags: |
+            type=ref,event=branch
+            type=ref,event=pr
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=semver,pattern={{major}}
+            type=sha
+            # set latest tag for default branch
+            type=raw,value=latest,enable=${{ github.event_name == 'release' }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Login to Docker Hub
+        if: github.event_name != 'pull_request' && env.DOCKERHUB_IMAGE_NAME != ''
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Login to GHCR
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          build-args: |
+            PYTORCH_INSTALL_ARGS=${{ matrix.pytorch_install_args }}
+            EXTRA_ARGS=${{ matrix.extra_args }}
+          cache-from: type=gha,scope=${{ github.ref_name }}-${{ matrix.id }}
+          cache-to: type=gha,mode=max,scope=${{ github.ref_name }}-${{ matrix.id }}
diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 000000000..ee361f276
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,79 @@
+# syntax=docker/dockerfile:1
+
+ARG PYTHON_VERSION=3.12
+
+FROM python:${PYTHON_VERSION}-slim
+
+ARG PYTORCH_INSTALL_ARGS=""
+ARG EXTRA_ARGS=""
+ARG USERNAME=comfyui
+ARG USER_UID=1000
+ARG USER_GID=${USER_UID}
+
+# Fail fast on errors or unset variables
+SHELL ["/bin/bash", "-eux", "-o", "pipefail", "-c"]
+
+RUN <<EOF
+	groupadd --gid ${USER_GID} ${USERNAME}
+	useradd --uid ${USER_UID} --gid ${USER_GID} -m ${USERNAME}
+EOF
+
+RUN <<EOF
+	apt-get update
+	apt-get install -y --no-install-recommends \
+		git \
+		git-lfs \
+		rsync \
+		fonts-recommended
+EOF
+
+# run instructions as user
+USER ${USER_UID}:${USER_GID}
+
+WORKDIR /app
+
+ENV XDG_CACHE_HOME=/cache
+ENV PIP_CACHE_DIR=/cache/pip
+ENV VIRTUAL_ENV=/app/venv
+ENV VIRTUAL_ENV_CUSTOM=/app/custom_venv
+
+# create cache directory. During build we will use a cache mount,
+# but later this is useful for custom node installs
+RUN --mount=type=cache,target=/cache/,uid=${USER_UID},gid=${USER_GID} \
+	mkdir -p ${PIP_CACHE_DIR}
+
+# create virtual environment to manage packages
+RUN python -m venv ${VIRTUAL_ENV}
+
+# run python from venv (prefer custom_venv over baked-in one)
+ENV PATH="${VIRTUAL_ENV_CUSTOM}/bin:${VIRTUAL_ENV}/bin:${PATH}"
+
+RUN --mount=type=cache,target=/cache/,uid=${USER_UID},gid=${USER_GID} \
+	pip install torch torchvision torchaudio ${PYTORCH_INSTALL_ARGS}
+
+# copy requirements files first so packages can be cached separately
+COPY --chown=${USER_UID}:${USER_GID} requirements.txt .
+RUN --mount=type=cache,target=/cache/,uid=${USER_UID},gid=${USER_GID} \
+	pip install -r requirements.txt
+
+# Not strictly required for comfyui, but prevents non-working variants of
+# cv2 being pulled in by custom nodes
+RUN --mount=type=cache,target=/cache/,uid=${USER_UID},gid=${USER_GID} \
+	pip install opencv-python-headless
+
+COPY --chown=${USER_UID}:${USER_GID} . .
+
+COPY --chown=nobody:${USER_GID} .git .git
+
+# default environment variables
+ENV COMFYUI_ADDRESS=0.0.0.0
+ENV COMFYUI_PORT=8188
+ENV COMFYUI_EXTRA_BUILD_ARGS="${EXTRA_ARGS}"
+ENV COMFYUI_EXTRA_ARGS=""
+# default start command
+CMD \
+	if [ -d "${VIRTUAL_ENV_CUSTOM}" ]; then \
+		rsync -aP "${VIRTUAL_ENV}/" "${VIRTUAL_ENV_CUSTOM}/" ;\
+		sed -i "s!${VIRTUAL_ENV}!${VIRTUAL_ENV_CUSTOM}!g" "${VIRTUAL_ENV_CUSTOM}/pyvenv.cfg" ;\
+	fi ;\
+	python -u main.py --listen "${COMFYUI_ADDRESS}" --port "${COMFYUI_PORT}" ${COMFYUI_EXTRA_BUILD_ARGS} ${COMFYUI_EXTRA_ARGS}
diff --git a/README.md b/README.md
index a99aca0e7..6798507b7 100644
--- a/README.md
+++ b/README.md
@@ -237,6 +237,45 @@ Install the dependencies by opening your terminal inside the ComfyUI folder and:
 
 After this you should have everything installed and can proceed to running ComfyUI.
 
+
+## Docker
+
+There are prebuilt docker images for AMD and NVIDIA GPUs on [GitHub Packages](https://ghcr.io/comfyanonymous/comfyui).
+
+You can pull them to your local docker registry with:
+
+```shell
+# For NVIDIA GPUs
+docker pull ghcr.io/comfyanonymous/comfyui:latest-cu124
+# For AMD GPUs
+docker pull ghcr.io/comfyanonymous/comfyui:latest-rocm6.2
+
+# For CPU only
+docker pull ghcr.io/comfyanonymous/comfyui:latest-cpu
+```
+
+### Building images manually
+
+You can build a docker image with the `Dockerfile` in this repo.
+Specify, `PYTORCH_INSTALL_ARGS` build arg with one of the PyTorch commands above to build for AMD or NVIDIA GPUs.
+```shell
+docker build --build-arg PYTORCH_INSTALL_ARGS="--index-url https://download.pytorch.org/whl/cu122" .
+```
+```shell
+docker build --build-arg PYTORCH_INSTALL_ARGS="--index-url https://download.pytorch.org/whl/rocm6.2" .
+```
+The `Dockerfile` requires BuildKit to be enabled. If your Docker does not support the buildx command, you can
+enable BuildKit by setting the `DOCKER_BUILDKIT` environment variable.
+```shell
+DOCKER_BUILDKIT=1 docker build --build-arg PYTORCH_INSTALL_ARGS="--index-url https://download.pytorch.org/whl/cu124" .
+```
+> [!NOTE]
+> For building the CPU-only image, it is recommended that you add the --cpu flag to the EXTRA_ARGS build arg:
+>
+> ```shell
+> docker build --build-arg PYTORCH_INSTALL_ARGS="--index-url https://download.pytorch.org/whl/cpu" --build-arg EXTRA_ARGS=--cpu .
+> ```
+
 ### Others:
 
 #### Apple Mac silicon
diff --git a/docker-compose.yaml b/docker-compose.yaml
new file mode 100644
index 000000000..858840fbe
--- /dev/null
+++ b/docker-compose.yaml
@@ -0,0 +1,21 @@
+services:
+  comfyui:
+    user: "1000:1000"
+    build: .
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
+    ports:
+      - "8188:8188"
+    volumes:
+      - "./models:/app/models"
+      - "./input:/app/input"
+      - "./temp:/app/output/temp"
+      - "./output:/app/output"
+      - "./user:/app/user"
+      - "./custom_venv:/app/custom_venv"
+      - "./custom_nodes:/app/custom_nodes"