diff --git a/.dockerignore b/.dockerignore
index 744da3a..c77a2dc 100644
--- a/.dockerignore
+++ b/.dockerignore
@@ -5,9 +5,10 @@
 .git
 .idea/
 __pycache__/
-tts/models/Wav2Lip/python/
-tts/models/tacotron2/python/
-tts/python/
+python/
 secrets.toml
 secrets.yaml
 target/
+tts/models/Wav2Lip/python/
+tts/models/tacotron2/python/
+tts/python/
diff --git a/Dockerfile b/Dockerfile
index 039d55a..ac472d2 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -230,6 +230,7 @@ LABEL org.opencontainers.image.authors='bt@brand.io, echelon@gmail.com'
 LABEL org.opencontainers.image.url='https://github.com/storytold/tts-core'
 LABEL org.opencontainers.image.documentation='https://github.com/storytold/tts-core'
 LABEL org.opencontainers.image.source='https://github.com/storytold/tts-core'
+LABEL org.opencontainers.image.description='Storyteller ML model monorepo: TTS, etc.'
 
 
 # NB(echelon,2022-07-06): The following fix for revoked keys was copied from "storyteller-rust".
@@ -277,5 +278,9 @@ RUN pip3 install gdown
 #  Fixes bug, "NVIDIA GeForce RTX 3080 with CUDA capability sm_86 is not compatible with the current PyTorch installation."
 #RUN pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
 
+# Mostly for local development:
+COPY ./start_tts_server.sh /
 
+# TTS model code:
 COPY ./tts ./models/tts
+
diff --git a/start_tts_server.sh b/start_tts_server.sh
new file mode 100755
index 0000000..f7d9339
--- /dev/null
+++ b/start_tts_server.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+
+cd models/tts
+source python/bin/activate
+./vocodes_server.py
+