diff --git a/Dockerfile b/Dockerfile
index 2452638fbd0c496c8574808058d9d23172f3c248..661bf92f020f36c0c8e29862640650f33ea95a94 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,11 +1,21 @@
 FROM python:3.8
 
+ARG with_models=false
+
 WORKDIR /app
 
 RUN pip install --upgrade pip
 
 COPY . .
 
+# check for offline build
+RUN if [ "$with_models" = "true" ]; then  \
+        # install only the dependencies first
+        pip install -e .;  \
+        # initialize the language models
+        ./install_models.py;  \
+    fi
+
 # Install package from source code
 RUN pip install .
 
diff --git a/README.md b/README.md
index 1ecd52ed38e0782449af18fd7f87446cf54eaaf2..706a751937a34d2b5fc4876acc2246da0a5c4849 100644
--- a/README.md
+++ b/README.md
@@ -80,9 +80,11 @@ Then open a web browser to http://localhost:5000
 ### Build with Docker
 
 ```bash
-docker build -t libretranslate .
+docker build [--build-arg with_models=true] -t libretranslate .
 ```
 
+If you want to run the Docker image in a complete offline environment, you need to add the `--build-arg with_models=true` parameter. Then the language models get downloaded during the build process of the image. Otherwise these models get downloaded on the first run of the image/container.
+
 Run the built image:
 
 ```bash