refactor(text2video): enhance video generation with audio and filter options

app/nlp.py

@@ -115,19 +115,53 @@ def generate_video(audio, text, filter_option):
         except Exception as e:
             return None, f"Error: {str(e)}"
 
+def call_grpc_server(audio, text, filter_option):
+    with grpc.insecure_channel('localhost:50051') as channel:
+        stub = text2video_pb2_grpc.VideoGeneratorStub(channel)
+
+        request = text2video_pb2.VideoRequest(
+            prompt=text.strip() if text else "",
+            audio_path=audio if audio else "",
+            filter_option=filter_option
+        )
+
+        response = stub.Generate(request)
+        if response.status_code == 200:
+            return response.video_path, f"<div>{response.message}</div>"
+        else:
+            return None, f"<div>{response.message}</div>"
+
 def serve():
     class VideoGeneratorServicer(text2video_pb2_grpc.VideoGeneratorServicer):
         def Generate(self, request, context):
+            # Extract and sanitize fields
             prompt = request.prompt.strip()
-            if not prompt:
+            audio_path = request.audio_path.strip()
+            filter_option = request.filter_option.strip()
+            print("---- Incoming Request ----")
+            print("Prompt:", request.prompt)
+            print("Audio Path:", request.audio_path)
+            print("Filter Option:", request.filter_option)
+
+            # Validate filter option
+            valid_filters = {"None", "Grayscale", "Sepia"}
+            if filter_option not in valid_filters:
+                return text2video_pb2.VideoResponse(
+                    video_path="",
+                    message=f"Invalid filter_option '{filter_option}'. Must be one of: {', '.join(valid_filters)}.",
+                    status_code=400
+                )
+
+            # Ensure at least one form of input
+            if not prompt and not audio_path:
                 return text2video_pb2.VideoResponse(
                     video_path="",
-                    message="Prompt cannot be empty.",
+                    message="Either prompt or audio path must be provided.",
                     status_code=400
                 )
 
             try:
-                video_path, _ = generate_video(prompt)
+                video_path, _ = generate_video(audio_path, prompt, filter_option)
                 if video_path and os.path.exists(video_path):
                     return text2video_pb2.VideoResponse(
                         video_path=video_path,
@@ -147,7 +181,7 @@ def serve():
                     status_code=500
                 )
 
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=4))
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=8))
     text2video_pb2_grpc.add_VideoGeneratorServicer_to_server(VideoGeneratorServicer(), server)
     server.add_insecure_port('[::]:50051')
     server.start()
@@ -155,7 +189,7 @@ def serve():
     server.wait_for_termination()
 
 iface = gr.Interface(
-    fn=generate_video,
+    fn=call_grpc_server,
     inputs=[
         gr.Audio(type="filepath", label="🎙 Upload or Record Audio (Optional)"),
         gr.Textbox(label="📝 Enter Text Prompt (Required if no audio)", placeholder="e.g., A knight fighting a dragon in the clouds"),

app/text2video.proto

@@ -2,19 +2,18 @@ syntax = "proto3";
 
 package text2video;
 
-// Request message
-message TextPrompt {
+message VideoRequest {
   string prompt = 1;
+  string audio_path = 2;
+  string filter_option = 3;
 }
 
-// Response message
 message VideoResponse {
   string video_path = 1;
   string message = 2;
   int32 status_code = 3;
 }
 
-// Service definition
 service VideoGenerator {
-  rpc Generate (TextPrompt) returns (VideoResponse);
+  rpc Generate (VideoRequest) returns (VideoResponse);
 }

app/text2video_pb2.py

@@ -24,17 +24,17 @@ _sym_db = _symbol_database.Default()
 
 
 
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x10text2video.proto\x12\ntext2video\"\x1c\n\nTextPrompt\x12\x0e\n\x06prompt\x18\x01 \x01(\t\"I\n\rVideoResponse\x12\x12\n\nvideo_path\x18\x01 \x01(\t\x12\x0f\n\x07message\x18\x02 \x01(\t\x12\x13\n\x0bstatus_code\x18\x03 \x01(\x05\x32O\n\x0eVideoGenerator\x12=\n\x08Generate\x12\x16.text2video.TextPrompt\x1a\x19.text2video.VideoResponseb\x06proto3')
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x10text2video.proto\x12\ntext2video\"I\n\x0cVideoRequest\x12\x0e\n\x06prompt\x18\x01 \x01(\t\x12\x12\n\naudio_path\x18\x02 \x01(\t\x12\x15\n\rfilter_option\x18\x03 \x01(\t\"I\n\rVideoResponse\x12\x12\n\nvideo_path\x18\x01 \x01(\t\x12\x0f\n\x07message\x18\x02 \x01(\t\x12\x13\n\x0bstatus_code\x18\x03 \x01(\x05\x32Q\n\x0eVideoGenerator\x12?\n\x08Generate\x12\x18.text2video.VideoRequest\x1a\x19.text2video.VideoResponseb\x06proto3')
 
 _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
 _builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'text2video_pb2', _globals)
 if not _descriptor._USE_C_DESCRIPTORS:
   DESCRIPTOR._loaded_options = None
-  _globals['_TEXTPROMPT']._serialized_start=32
-  _globals['_TEXTPROMPT']._serialized_end=60
-  _globals['_VIDEORESPONSE']._serialized_start=62
-  _globals['_VIDEORESPONSE']._serialized_end=135
-  _globals['_VIDEOGENERATOR']._serialized_start=137
-  _globals['_VIDEOGENERATOR']._serialized_end=216
+  _globals['_VIDEOREQUEST']._serialized_start=32
+  _globals['_VIDEOREQUEST']._serialized_end=105
+  _globals['_VIDEORESPONSE']._serialized_start=107
+  _globals['_VIDEORESPONSE']._serialized_end=180
+  _globals['_VIDEOGENERATOR']._serialized_start=182
+  _globals['_VIDEOGENERATOR']._serialized_end=263
 # @@protoc_insertion_point(module_scope)

app/text2video_pb2_grpc.py

@@ -26,8 +26,7 @@ if _version_not_supported:
 
 
 class VideoGeneratorStub(object):
-    """Service definition
-    """
+    """Missing associated documentation comment in .proto file."""
 
     def __init__(self, channel):
         """Constructor.
@@ -37,14 +36,13 @@ class VideoGeneratorStub(object):
         """
         self.Generate = channel.unary_unary(
                 '/text2video.VideoGenerator/Generate',
-                request_serializer=text2video__pb2.TextPrompt.SerializeToString,
+                request_serializer=text2video__pb2.VideoRequest.SerializeToString,
                 response_deserializer=text2video__pb2.VideoResponse.FromString,
                 _registered_method=True)
 
 
 class VideoGeneratorServicer(object):
-    """Service definition
-    """
+    """Missing associated documentation comment in .proto file."""
 
     def Generate(self, request, context):
         """Missing associated documentation comment in .proto file."""
@@ -57,7 +55,7 @@ def add_VideoGeneratorServicer_to_server(servicer, server):
     rpc_method_handlers = {
             'Generate': grpc.unary_unary_rpc_method_handler(
                     servicer.Generate,
-                    request_deserializer=text2video__pb2.TextPrompt.FromString,
+                    request_deserializer=text2video__pb2.VideoRequest.FromString,
                     response_serializer=text2video__pb2.VideoResponse.SerializeToString,
             ),
     }
@@ -69,8 +67,7 @@ def add_VideoGeneratorServicer_to_server(servicer, server):
 
  # This class is part of an EXPERIMENTAL API.
 class VideoGenerator(object):
-    """Service definition
-    """
+    """Missing associated documentation comment in .proto file."""
 
     @staticmethod
     def Generate(request,
@@ -87,7 +84,7 @@ class VideoGenerator(object):
             request,
             target,
             '/text2video.VideoGenerator/Generate',
-            text2video__pb2.TextPrompt.SerializeToString,
+            text2video__pb2.VideoRequest.SerializeToString,
             text2video__pb2.VideoResponse.FromString,
             options,
             channel_credentials,