support s3

src/predict.py

@@ -58,6 +58,7 @@ from pydantic import BaseModel, HttpUrl
 from api import BaseSpeakerTTS, ToneColorConverter
 
 from pydub import AudioSegment
+from urllib.parse import urlparse
 
 
 class Predictor:
@@ -146,22 +147,32 @@ class Predictor:
 
 
     def createvoice(self,audio_base_64,cut_audio,process_audio):
-        file_bytes = base64.b64decode(audio_base_64)
-        file_buffer = io.BytesIO(file_bytes)
+        file_bytes = None
+        if s3_url:
+            parsed_url = urlparse(s3_url)
+            bucket_name = parsed_url.netloc
+            s3_key = parsed_url.path.lstrip('/')
+            local_filename = f"{uuid.uuid4()}"
+            self.download_file_from_s3(bucket_name, s3_key, local_filename)
+            with open(local_filename, 'rb') as file:
+                file_bytes = file.read()
+            os.remove(local_filename) 
+        elif audio_base_64:
+            file_bytes = base64.b64decode(audio_base_64)
+        else:
+            raise ValueError("Either s3_url or audio_base_64 must be provided.")
 
+        file_buffer = io.BytesIO(file_bytes)
         header = file_buffer.read(12)
         print(header)
-        file_format = None
-        bucket_name = self.bucket_name
+
+        file_format = 'mp3'  # Default format
         if b'WAVE' in header:
             file_format = 'wav'
         elif b'OggS' in header:
             file_format = 'ogg'
-        else:
-            file_format = 'mp3'
 
         unique_filename = f"{uuid.uuid4()}"
-
         local_filename = f"{unique_filename}.{file_format}"
         with open(local_filename, 'wb') as file_out:
             file_out.write(file_bytes)

src/rp_handler.py

@@ -28,9 +28,10 @@ def run_voice_clone_job(job):
         return {"error":"Please set method_type: available options, create_voice, voice_clone, voice_clone_with_emotions,voice_clone_with_multi_lang"}
     
     if method_type == "create_voice":
+        s3_url = job_input.get("s3_url")
         audio_base64 = job_input.get('audio_base64')
-        if audio_base64 is None:
-            return {"error":"Needs audio file as base64"}
+        if audio_base64 is None and s3_url is None:
+            return {"error":"set audio_base64 or s3_url"}
         cut_audio = job_input.get('cut_audio')
         process_audio = job_input.get('process_audio')
         print(process_audio)
@@ -39,7 +40,7 @@ def run_voice_clone_job(job):
         if cut_audio is None:
             cut_audio = 0
         
-        processed_urls = MODEL.createvoice(audio_base64,cut_audio,process_audio)
+        processed_urls = MODEL.createvoice(s3_url,audio_base64,cut_audio,process_audio)
         return processed_urls
     else:
         s3_url = job_input.get('s3_url')