diff --git a/audiotools/core/audio_signal.py b/audiotools/core/audio_signal.py
index fb6d751..7adb822 100644
--- a/audiotools/core/audio_signal.py
+++ b/audiotools/core/audio_signal.py
@@ -17,14 +17,12 @@
 from . import util
 from .display import DisplayMixin
 from .dsp import DSPMixin
-from .effects import EffectMixin
-from .effects import ImpulseResponseMixin
+from .effects import EffectMixin, ImpulseResponseMixin
 from .ffmpeg import FFMPEGMixin
 from .loudness import LoudnessMixin
 from .playback import PlayMixin
 from .whisper import WhisperMixin
 
-
 STFTParams = namedtuple(
     "STFTParams",
     ["window_length", "hop_length", "window_type", "match_stride", "padding_type"],
@@ -88,7 +86,8 @@ class AudioSignal(
     duration : float, optional
         Duration in seconds to read from file, by default None
     device : str, optional
-        Device to load audio onto, by default None
+        Device to load audio onto. The default for files in CPU, and default for
+        tensors and numpy arrays is the same device that they are on.
 
     Examples
     --------
@@ -155,6 +154,8 @@ def __init__(
                 audio_path, offset=offset, duration=duration, device=device
             )
         elif audio_array is not None:
+            if device is None:
+                device = audio_array.device
             assert sample_rate is not None, "Must set sample rate!"
             self.load_from_array(audio_array, sample_rate, device=device)