Update app.py

app.py

@@ -35,9 +35,14 @@ def tts(text,lang,speaker,
             model.generate(model.prepare_conditioning(cond),
                            cfg_scale=float(cfg),sampling_params=dict(min_p=float(minp)))
         )[0].cpu().clamp_(-1,1).numpy()
+    wav = np.squeeze(wav)                    # drop singleton channel
+    if wav.ndim == 2:                        # (C, N) ➜ (N, C)
+        wav = wav.T
+    
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
-        sf.write(tmp.name, wav, model.autoencoder.sampling_rate,
-            subtype="PCM_16")
+        sf.write(tmp.name, wav,
+                 model.autoencoder.sampling_rate,
+                 format="WAV", subtype="PCM_16")
     return tmp.name
 
 langs=supported_language_codes