Fixed float type error.

chatterbox/tts_turbo.py

@@ -233,6 +233,8 @@ class ChatterboxTurboTTS:
             s3gen_ref_wav = self.norm_loudness(s3gen_ref_wav, _sr)
 
         ref_16k_wav = librosa.resample(s3gen_ref_wav, orig_sr=S3GEN_SR, target_sr=S3_SR)
+        ref_16k_wav = ref_16k_wav.astype("float32")  # librosa.resample returns float64
+
 
         s3gen_ref_wav = s3gen_ref_wav[:self.DEC_COND_LEN]
         s3gen_ref_dict = self.s3gen.embed_ref(s3gen_ref_wav, S3GEN_SR, device=self.device)