diff --git a/docs/docs/install/environment-variables.md b/docs/docs/install/environment-variables.md
index 78cd16cf1b..9a4b0b9360 100644
--- a/docs/docs/install/environment-variables.md
+++ b/docs/docs/install/environment-variables.md
@@ -159,26 +159,29 @@ Redis (Sentinel) URL example JSON before encoding:
## Machine Learning
-| Variable | Description | Default | Containers |
-| :----------------------------------------------- | :-------------------------------------------------------------------------------------------------- | :-----------------------------------: | :--------------- |
-| `MACHINE_LEARNING_MODEL_TTL` | Inactivity time (s) before a model is unloaded (disabled if \<= 0) | `300` | machine learning |
-| `MACHINE_LEARNING_MODEL_TTL_POLL_S` | Interval (s) between checks for the model TTL (disabled if \<= 0) | `10` | machine learning |
-| `MACHINE_LEARNING_CACHE_FOLDER` | Directory where models are downloaded | `/cache` | machine learning |
-| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0) | number of CPU cores | machine learning |
-| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS` | Number of parallel model operations | `1` | machine learning |
-| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS` | Number of threads for each model operation | `2` | machine learning |
-| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup> | Number of worker processes to spawn | `1` | machine learning |
-| `MACHINE_LEARNING_WORKER_TIMEOUT` | Maximum time (s) of unresponsiveness before a worker is killed | `120` (`300` if using OpenVINO image) | machine learning |
-| `MACHINE_LEARNING_PRELOAD__CLIP` | Name of a CLIP model to be preloaded and kept in cache | | machine learning |
-| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION` | Name of a facial recognition model to be preloaded and kept in cache | | machine learning |
-| `MACHINE_LEARNING_ANN` | Enable ARM-NN hardware acceleration if supported | `True` | machine learning |
-| `MACHINE_LEARNING_ANN_FP16_TURBO` | Execute operations in FP16 precision: increasing speed, reducing precision (applies only to ARM-NN) | `False` | machine learning |
-| `MACHINE_LEARNING_ANN_TUNING_LEVEL` | ARM-NN GPU tuning level (1: rapid, 2: normal, 3: exhaustive) | `2` | machine learning |
+| Variable | Description | Default | Containers |
+| :-------------------------------------------------------- | :-------------------------------------------------------------------------------------------------- | :-----------------------------------: | :--------------- |
+| `MACHINE_LEARNING_MODEL_TTL` | Inactivity time (s) before a model is unloaded (disabled if \<= 0) | `300` | machine learning |
+| `MACHINE_LEARNING_MODEL_TTL_POLL_S` | Interval (s) between checks for the model TTL (disabled if \<= 0) | `10` | machine learning |
+| `MACHINE_LEARNING_CACHE_FOLDER` | Directory where models are downloaded | `/cache` | machine learning |
+| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0) | number of CPU cores | machine learning |
+| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS` | Number of parallel model operations | `1` | machine learning |
+| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS` | Number of threads for each model operation | `2` | machine learning |
+| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup> | Number of worker processes to spawn | `1` | machine learning |
+| `MACHINE_LEARNING_HTTP_KEEPALIVE_TIMEOUT_S`<sup>\*3</sup> | HTTP Keep-alive time in seconds | `2` | machine learning |
+| `MACHINE_LEARNING_WORKER_TIMEOUT` | Maximum time (s) of unresponsiveness before a worker is killed | `120` (`300` if using OpenVINO image) | machine learning |
+| `MACHINE_LEARNING_PRELOAD__CLIP` | Name of a CLIP model to be preloaded and kept in cache | | machine learning |
+| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION` | Name of a facial recognition model to be preloaded and kept in cache | | machine learning |
+| `MACHINE_LEARNING_ANN` | Enable ARM-NN hardware acceleration if supported | `True` | machine learning |
+| `MACHINE_LEARNING_ANN_FP16_TURBO` | Execute operations in FP16 precision: increasing speed, reducing precision (applies only to ARM-NN) | `False` | machine learning |
+| `MACHINE_LEARNING_ANN_TUNING_LEVEL` | ARM-NN GPU tuning level (1: rapid, 2: normal, 3: exhaustive) | `2` | machine learning |
\*1: It is recommended to begin with this parameter when changing the concurrency levels of the machine learning service and then tune the other ones.
\*2: Since each process duplicates models in memory, changing this is not recommended unless you have abundant memory to go around.
+\*3: For scenarios like HPA in K8S. https://github.com/immich-app/immich/discussions/12064
+
:::info
Other machine learning parameters can be tuned from the admin UI.
diff --git a/machine-learning/start.sh b/machine-learning/start.sh
index 6b8e55a236..c3fda523df 100755
--- a/machine-learning/start.sh
+++ b/machine-learning/start.sh
@@ -13,6 +13,7 @@ fi
: "${IMMICH_HOST:=[::]}"
: "${IMMICH_PORT:=3003}"
: "${MACHINE_LEARNING_WORKERS:=1}"
+: "${MACHINE_LEARNING_HTTP_KEEPALIVE_TIMEOUT_S:=2}"
gunicorn app.main:app \
-k app.config.CustomUvicornWorker \
@@ -20,4 +21,5 @@ gunicorn app.main:app \
-w "$MACHINE_LEARNING_WORKERS" \
-t "$MACHINE_LEARNING_WORKER_TIMEOUT" \
--log-config-json log_conf.json \
+ --keep-alive "$MACHINE_LEARNING_HTTP_KEEPALIVE_TIMEOUT_S" \
--graceful-timeout 0