Add NVIDIA support for FFmpeg in Docker and enhance chat rendering functionality

- Introduced a new docker-compose.nvidia.yml for NVIDIA GPU support.
- Updated dockerstart.bat to allow optional NVIDIA runtime.
- Enhanced ContentDownloader to manage chat rendering status and font settings.
- Improved hardware acceleration detection in utils.py.
- Added tests for hardware acceleration and chat rendering behavior.

Co-authored-by: Copilot <copilot@github.com>

modules/utils.py

@@ -269,20 +269,57 @@ def detect_hardware_acceleration(hwaccel_config: str, os_type: str) -> Optional[
     if hwaccel_config in ['nvenc', 'qsv', 'amf', 'vaapi']:
         return hwaccel_config
     
-    # Auto-detect: try to determine available hardware
+    # Auto-detect: choose only hardware we can reasonably prove is present.
     if hwaccel_config == 'auto':
-        # On Windows, NVIDIA is most common
-        if os_type == 'windows':
-            # Could check for nvidia-smi, but just return 'auto' for ffmpeg to decide
-            return 'auto'
-        else:
-            # On Linux, VAAPI is common for Intel/AMD, or NVENC for NVIDIA
-            # Let ffmpeg auto-detect
-            return 'auto'
+        if is_nvidia_runtime_available():
+            return 'nvenc'
+        if is_vaapi_runtime_available():
+            return 'vaapi'
+        return 'none'
     
     return None
 
 
+def is_nvidia_runtime_available() -> bool:
+    """Return True when the current runtime appears to expose an NVIDIA GPU."""
+    visible_devices = os.getenv('NVIDIA_VISIBLE_DEVICES', '').strip().lower()
+    if visible_devices in {'void', 'none'}:
+        return False
+    if visible_devices and visible_devices != 'all':
+        return True
+
+    if shutil.which('nvidia-smi'):
+        return True
+
+    return any(
+        os.path.exists(device_path)
+        for device_path in ('/dev/nvidiactl', '/dev/nvidia0', '/dev/nvidia-modeset')
+    )
+
+
+def is_vaapi_runtime_available() -> bool:
+    """Return True when Linux VAAPI render nodes are present."""
+    return any(
+        os.path.exists(device_path)
+        for device_path in ('/dev/dri/renderD128', '/dev/dri/card0')
+    )
+
+
+def resolve_hwaccel_type(hwaccel_type: Optional[str], os_type: str) -> Optional[str]:
+    """Return a safe hardware acceleration choice for the current runtime."""
+    if hwaccel_type in (None, 'none'):
+        return 'none'
+
+    if hwaccel_type == 'nvenc':
+        return 'nvenc' if is_nvidia_runtime_available() else 'none'
+
+    if hwaccel_type == 'vaapi':
+        return 'vaapi' if is_vaapi_runtime_available() else 'none'
+
+    # Leave explicit QSV/AMF unchanged for non-container users; container auto-detect no longer picks them blindly.
+    return hwaccel_type
+
+
 def get_hwaccel_encoder(hwaccel_type: str) -> str:
     """
     Get the appropriate hardware-accelerated encoder for the given acceleration type.