diff --git a/README.md b/README.md
index eb90b4d..6859163 100644
--- a/README.md
+++ b/README.md
@@ -91,11 +91,19 @@ docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:v0.2.2  #NV
         Run the [model download script](https://github.com/remsky/Kokoro-FastAPI/blob/master/docker/scripts/download_model.py) if you haven't already
      
         Start directly via UV (with hot-reload)
+        
+        Linux and macOS
         ```bash
         ./start-cpu.sh OR
         ./start-gpu.sh 
         ```
 
+        Windows
+        ```powershell
+        .\start-cpu.ps1 OR
+        .\start-gpu.ps1 
+        ```
+
 </details>
 
 <details open>
diff --git a/start-cpu.ps1 b/start-cpu.ps1
new file mode 100644
index 0000000..5a5df26
--- /dev/null
+++ b/start-cpu.ps1
@@ -0,0 +1,13 @@
+$env:PHONEMIZER_ESPEAK_LIBRARY="C:\Program Files\eSpeak NG\libespeak-ng.dll"
+$env:PYTHONUTF8=1
+$Env:PROJECT_ROOT="$pwd"
+$Env:USE_GPU="false"
+$Env:USE_ONNX="false"
+$Env:PYTHONPATH="$Env:PROJECT_ROOT;$Env:PROJECT_ROOT/api"
+$Env:MODEL_DIR="src/models"
+$Env:VOICES_DIR="src/voices/v1_0"
+$Env:WEB_PLAYER_PATH="$Env:PROJECT_ROOT/web"
+
+uv pip install -e ".[cpu]"
+uv run --no-sync python docker/scripts/download_model.py --output api/src/models/v1_0
+uv run --no-sync uvicorn api.src.main:app --host 0.0.0.0 --port 8880
\ No newline at end of file
diff --git a/start-cpu.sh b/start-cpu.sh
index 651f645..4438ff9 100755
--- a/start-cpu.sh
+++ b/start-cpu.sh
@@ -14,4 +14,5 @@ export WEB_PLAYER_PATH=$PROJECT_ROOT/web
 # Run FastAPI with CPU extras using uv run
 # Note: espeak may still require manual installation,
 uv pip install -e ".[cpu]"
+uv run --no-sync python docker/scripts/download_model.py --output api/src/models/v1_0
 uv run --no-sync uvicorn api.src.main:app --host 0.0.0.0 --port 8880
diff --git a/start-gpu.ps1 b/start-gpu.ps1
new file mode 100644
index 0000000..7b161a5
--- /dev/null
+++ b/start-gpu.ps1
@@ -0,0 +1,13 @@
+$env:PHONEMIZER_ESPEAK_LIBRARY="C:\Program Files\eSpeak NG\libespeak-ng.dll"
+$env:PYTHONUTF8=1
+$Env:PROJECT_ROOT="$pwd"
+$Env:USE_GPU="true"
+$Env:USE_ONNX="false"
+$Env:PYTHONPATH="$Env:PROJECT_ROOT;$Env:PROJECT_ROOT/api"
+$Env:MODEL_DIR="src/models"
+$Env:VOICES_DIR="src/voices/v1_0"
+$Env:WEB_PLAYER_PATH="$Env:PROJECT_ROOT/web"
+
+uv pip install -e ".[gpu]"
+uv run --no-sync python docker/scripts/download_model.py --output api/src/models/v1_0
+uv run --no-sync uvicorn api.src.main:app --host 0.0.0.0 --port 8880
\ No newline at end of file
diff --git a/start-gpu.sh b/start-gpu.sh
index b079978..a3a2e68 100755
--- a/start-gpu.sh
+++ b/start-gpu.sh
@@ -12,5 +12,7 @@ export VOICES_DIR=src/voices/v1_0
 export WEB_PLAYER_PATH=$PROJECT_ROOT/web
 
 # Run FastAPI with GPU extras using uv run
+# Note: espeak may still require manual installation,
 uv pip install -e ".[gpu]"
+uv run --no-sync python docker/scripts/download_model.py --output api/src/models/v1_0
 uv run --no-sync uvicorn api.src.main:app --host 0.0.0.0 --port 8880