diff --git a/tortoise/api.py b/tortoise/api.py
index 296ef14a..088c7250 100644
--- a/tortoise/api.py
+++ b/tortoise/api.py
@@ -1,3 +1,4 @@
+import logging
 import os
 import random
 import uuid
@@ -213,7 +214,7 @@ def __init__(self, autoregressive_batch_size=None, models_dir=MODELS_DIR, enable
         self.models_dir = models_dir
         self.autoregressive_batch_size = pick_best_batch_size_for_gpu() if autoregressive_batch_size is None else autoregressive_batch_size
         self.enable_redaction = enable_redaction
-        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+        self._set_device(device)
         if self.enable_redaction:
             self.aligner = Wav2VecAlignment()
 
@@ -250,6 +251,14 @@ def __init__(self, autoregressive_batch_size=None, models_dir=MODELS_DIR, enable
         self.rlg_auto = None
         self.rlg_diffusion = None
 
+    def _set_device(self, device):
+        if device is None:
+            logging.warning("No device specified. This will default to the first GPU if available, otherwise CPU.")
+            device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+        if device == 'cpu' or (isinstance(device, torch.device) and device.type == 'cpu'):
+            logging.warning("Running on CPU. This will be painfully slow. You've been warned.")
+        self.device = device
+
     def load_cvvp(self):
         """Load CVVP model."""
         self.cvvp = CVVP(model_dim=512, transformer_heads=8, dropout=0, mel_codes=8192, conditioning_enc_depth=8, cond_mask_percentage=0,