Fix and enable XTTS streaming

script.py

@@ -3309,6 +3309,7 @@ def on_load(request: gr.Request):
                         gen_lang = gr.Dropdown(
                             value=config.api_def.api_language,
                             choices=[
+                                "auto",
                                 "ar",
                                 "zh",
                                 "cs",

system/requirements/requirements_colab.txt

@@ -46,3 +46,4 @@ piper-tts; sys_platform == "linux"
 plotly==5.24.1
 scipy==1.14.1
 pyOpenSSL>=24.2.1
+langdetect>=1.0.9

system/requirements/requirements_standalone.txt

@@ -36,3 +36,4 @@ fastapi==0.112.2
 plotly==5.24.1
 scipy==1.14.1
 pyOpenSSL>=24.2.1
+langdetect>=1.0.9

system/requirements/requirements_textgen.txt

@@ -33,3 +33,4 @@ piper-phonemize==1.1.0; sys_platform == "darwin"
 plotly==5.24.1
 scipy==1.14.1
 pyOpenSSL>=24.2.1
+langdetect>=1.0.9

system/tts_engines/f5tts/f5tts_settings_page.py

@@ -140,7 +140,7 @@ def transcribe_files(model_name: str, use_cpu: bool = False, progress=gr.Progres
 # dictionaries with the values provided as arguments, and save the updated settings back to the JSON file.
 #
 # You do not need to modify the function's logic or any other part of the code.
-def f5tts_model_update_settings(def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr,  alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr):
+def f5tts_model_update_settings(def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, streaming_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr,  alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr):
     # Load the model_config_data from the JSON file
     with open(os.path.join(this_dir, "model_settings.json"), "r") as f:
         model_config_data = json.load(f)
@@ -155,6 +155,7 @@ def f5tts_model_update_settings(def_character_voice_gr, def_narrator_voice_gr, l
     model_config_data["openai_voices"]["shimmer"] = shimmer_gr
     model_config_data["settings"]["lowvram_enabled"] = lowvram_enabled_gr == "Enabled"
     model_config_data["settings"]["deepspeed_enabled"] = deepspeed_enabled_gr == "Enabled"
+    model_config_data["settings"]["streaming_enabled"] = streaming_enabled_gr == "Enabled"
     model_config_data["settings"]["temperature_set"] = temperature_set_gr
     model_config_data["settings"]["repetitionpenalty_set"] = repetitionpenalty_set_gr
     model_config_data["settings"]["pitch_set"] = pitch_set_gr
@@ -192,6 +193,7 @@ def f5tts_model_alltalk_settings(model_config_data):
             with gr.Row():
                 lowvram_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="Low VRAM" if model_config_data["model_capabilties"]["lowvram_capable"] else "Low VRAM N/A", value="Enabled" if model_config_data["settings"]["lowvram_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["lowvram_capable"])
                 deepspeed_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="DeepSpeed Activate" if model_config_data["model_capabilties"]["deepspeed_capable"] else "DeepSpeed N/A", value="Enabled" if model_config_data["settings"]["deepspeed_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["deepspeed_capable"])
+                streaming_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="Streaming" if model_config_data["model_capabilties"]["streaming_capable"] else "Streaming N/A", value="Enabled" if model_config_data["settings"]["streaming_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["streaming_capable"])
                 temperature_set_gr = gr.Slider(value=float(model_config_data["settings"]["temperature_set"]), minimum=0, maximum=1, step=0.05, label="Temperature" if model_config_data["model_capabilties"]["temperature_capable"] else "Temperature N/A", interactive=model_config_data["model_capabilties"]["temperature_capable"])
                 repetitionpenalty_set_gr = gr.Slider(value=float(model_config_data["settings"]["repetitionpenalty_set"]), minimum=1, maximum=20, step=1, label="Repetition Penalty" if model_config_data["model_capabilties"]["repetitionpenalty_capable"] else "Repetition N/A", interactive=model_config_data["model_capabilties"]["repetitionpenalty_capable"])
                 pitch_set_gr = gr.Slider(value=float(model_config_data["settings"]["pitch_set"]), minimum=-10, maximum=10, step=1, label="Pitch" if model_config_data["model_capabilties"]["pitch_capable"] else "Pitch N/A", interactive=model_config_data["model_capabilties"]["pitch_capable"])
@@ -226,7 +228,7 @@ def f5tts_model_alltalk_settings(model_config_data):
                 with gr.Row():
                     gr.Markdown(AllTalkHelpContent.DEFAULT_SETTINGS1, elem_classes="custom-markdown")
                     gr.Markdown(AllTalkHelpContent.DEFAULT_SETTINGS2, elem_classes="custom-markdown")                
-            submit_button.click(f5tts_model_update_settings, inputs=[def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr, alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr], outputs=output_message)
+            submit_button.click(f5tts_model_update_settings, inputs=[def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, streaming_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr, alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr], outputs=output_message)
 
         ###########################################################################################
         # Do not change this section apart from "TTS Engine Name" value to match your engine name #

system/tts_engines/f5tts/help_content.py

@@ -237,6 +237,11 @@ class AllTalkHelpContent:
         - Accelerates TTS generation using optimized inference
         - Only available for engines and models that support DeepSpeed
         - Requires NVIDIA GPU with CUDA support
+
+    - **Stream Response Capability**
+        - Enables real-time streaming of generated speech output
+        - Reduces latency for faster feedback during synthesis
+        - Only available for engines and models that support Streaming
 
     - **Temperature Control**
         - Adjusts the variability in speech generation

system/tts_engines/f5tts/model_engine.py

@@ -144,6 +144,7 @@ def __init__(self):
         self.def_character_voice = tts_model_loaded["settings"]["def_character_voice"]                      # What is the current default main/character voice that will be used if no voice specified.
         self.def_narrator_voice = tts_model_loaded["settings"]["def_narrator_voice"]                        # What is the current default narrator voice that will be used if no voice specified.
         self.deepspeed_enabled = tts_model_loaded["settings"]["deepspeed_enabled"]                          # If its available, is DeepSpeed enabled for the TTS engine
+        self.streaming_enabled = tts_model_loaded["settings"]["streaming_enabled"]                          # If its available, is Streaming enabled for the TTS engine
         self.engine_installed = tts_model_loaded["settings"]["engine_installed"]                            # Has the TTS engine been setup/installed (not curently used)
         self.generationspeed_set = tts_model_loaded["settings"]["generationspeed_set"]                      # What is the set/stored speed for generation.
         self.lowvram_enabled = tts_model_loaded["settings"]["lowvram_enabled"]                              # If its available, is LowVRAM enabled for the TTS engine
@@ -400,9 +401,16 @@ def scan_models_folder(self):
             if model_dir.is_dir():
                 # First try to find model_*.safetensors files
                 model_files = list(model_dir.glob("model_*.safetensors"))
+                if not model_files:
+                # Try finding the pt model file as fallback
+                    # If no model_*.safetensors found, try finding a .pt model file
+                    model_files = list(model_dir.glob("model_*.pt"))
                 if not model_files:
                     # If no model_*.safetensors found, try any .safetensors file
                     model_files = list(model_dir.glob("*.safetensors"))
+                if not model_files:
+                    # If no model_*.safetensors found, try any .pt file
+                    model_files = list(model_dir.glob("*.pt"))
 
                 vocab_file = model_dir / "vocab.txt"
                 vocos_dir = model_dir / "vocos"
@@ -508,9 +516,15 @@ async def api_manual_load_model(self, model_name):
 
             # Dynamically find the safetensors model file
             model_files = list(model_dir.glob("model_*.safetensors"))
+            if not model_files:
+                # Try finding the pt model file as fallback
+                model_files = list(model_dir.glob("model_*.pt"))
             if not model_files:
                 # Try finding any safetensors file as fallback
                 model_files = list(model_dir.glob("*.safetensors"))
+            if not model_files:
+                # Try finding any pt file as fallback
+                model_files = list(model_dir.glob("*.pt"))
 
             if not model_files:
                 print(f"[{self.branding}ENG] \033[91mError\033[0m: No model's safetensors file was found in the F5-TTS models directory.")
@@ -1082,7 +1096,7 @@ async def generate_tts(self, text, voice, language, temperature, repetition_pena
 
             generate_end_time = time.time()
             generate_elapsed_time = generate_end_time - generate_start_time
-            print(f"[{self.branding}GEN] \033[94mTTS Generate: \033[93m{generate_elapsed_time:.2f} seconds. \033[94mLowVRAM: \033[33m{self.lowvram_enabled} \033[94mDeepSpeed: \033[33m{self.deepspeed_enabled}\033[0m")
+            print(f"[{self.branding}GEN] \033[94mTTS Generate: \033[93m{generate_elapsed_time:.2f} seconds. \033[94mLowVRAM: \033[33m{self.lowvram_enabled} \033[94mDeepSpeed: \033[33m{self.deepspeed_enabled} \033[94mStreaming: \033[33m{self.streaming_enabled}\033[0m")
 
             if streaming:
                 with open(output_file, 'rb') as f:

system/tts_engines/f5tts/model_settings.json

@@ -24,6 +24,7 @@
         "def_character_voice": "female_01.wav",
         "def_narrator_voice": "female_01.wav",
         "deepspeed_enabled": false,
+        "streaming_enabled": false,
         "engine_installed": true,
         "generationspeed_set": 0.9,
         "lowvram_enabled": true,

system/tts_engines/parler/help_content.py

@@ -237,6 +237,11 @@ class AllTalkHelpContent:
         - Accelerates TTS generation using optimized inference
         - Only available for engines and models that support DeepSpeed
         - Requires NVIDIA GPU with CUDA support
+
+    - **Stream Response Capability**
+        - Enables real-time streaming of generated speech output
+        - Reduces latency for faster feedback during synthesis
+        - Only available for engines and models that support Streaming
 
     - **Temperature Control**
         - Adjusts the variability in speech generation

system/tts_engines/parler/model_engine.py

@@ -91,6 +91,7 @@ def __init__(self):
         self.def_character_voice = tts_model_loaded["settings"]["def_character_voice"]                      # What is the current default main/character voice that will be used if no voice specified.
         self.def_narrator_voice = tts_model_loaded["settings"]["def_narrator_voice"]                        # What is the current default narrator voice that will be used if no voice specified.
         self.deepspeed_enabled = tts_model_loaded["settings"]["deepspeed_enabled"]                          # If its available, is DeepSpeed enabled for the TTS engine
+        self.streaming_enabled = tts_model_loaded["settings"]["streaming_enabled"]                          # If its available, is Streaming enabled for the TTS engine
         self.engine_installed = tts_model_loaded["settings"]["engine_installed"]                            # Has the TTS engine been setup/installed (not curently used)
         self.generationspeed_set = tts_model_loaded["settings"]["generationspeed_set"]                      # What is the set/stored speed for generation.
         self.lowvram_enabled = tts_model_loaded["settings"]["lowvram_enabled"]                              # If its available, is LowVRAM enabled for the TTS engine
@@ -503,7 +504,7 @@ async def generate_tts(self, text, voice, language, temperature, repetition_pena
         # ↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓ 
         generate_end_time = time.time()  # Record the end time to generate TTS
         generate_elapsed_time = generate_end_time - generate_start_time
-        print(f"[{self.branding}GEN] \033[94mTTS Generate: \033[93m{generate_elapsed_time:.2f} seconds. \033[94mLowVRAM: \033[33m{self.lowvram_enabled} \033[94mDeepSpeed: \033[33m{self.deepspeed_enabled}\033[0m")
+        print(f"[{self.branding}GEN] \033[94mTTS Generate: \033[93m{generate_elapsed_time:.2f} seconds. \033[94mLowVRAM: \033[33m{self.lowvram_enabled} \033[94mDeepSpeed: \033[33m{self.deepspeed_enabled} \033[94mStreaming: \033[33m{self.streaming_enabled}\033[0m")
         if self.lowvram_enabled and self.device == "cuda" and self.tts_narrator_generatingtts == False:
             await self.handle_lowvram_change()
         self.tts_generating_lock = False # Unlock the TTS generation queue to allow TTS generation requests to come in again.

system/tts_engines/parler/model_settings.json

@@ -24,6 +24,7 @@
         "def_character_voice": "enthusiastic_female",
         "def_narrator_voice": "enthusiastic_female",
         "deepspeed_enabled": false,
+        "streaming_enabled": false,
         "engine_installed": true,
         "generationspeed_set": 1,
         "lowvram_enabled": false,

system/tts_engines/parler/parler_settings_page.py

@@ -52,7 +52,7 @@ def parler_voices_file_list():
 # dictionaries with the values provided as arguments, and save the updated settings back to the JSON file.
 #
 # You do not need to modify the function's logic or any other part of the code.
-def parler_model_update_settings(def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr,  alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr):
+def parler_model_update_settings(def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, streaming_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr,  alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr):
     # Load the model_config_data from the JSON file
     with open(os.path.join(this_dir, "model_settings.json"), "r") as f:
         model_config_data = json.load(f)
@@ -67,6 +67,7 @@ def parler_model_update_settings(def_character_voice_gr, def_narrator_voice_gr,
     model_config_data["openai_voices"]["shimmer"] = shimmer_gr
     model_config_data["settings"]["lowvram_enabled"] = lowvram_enabled_gr == "Enabled"
     model_config_data["settings"]["deepspeed_enabled"] = deepspeed_enabled_gr == "Enabled"
+    model_config_data["settings"]["streaming_enabled"] = streaming_enabled_gr == "Enabled"
     model_config_data["settings"]["temperature_set"] = temperature_set_gr
     model_config_data["settings"]["repetitionpenalty_set"] = repetitionpenalty_set_gr
     model_config_data["settings"]["pitch_set"] = pitch_set_gr
@@ -104,6 +105,7 @@ def parler_model_alltalk_settings(model_config_data):
             with gr.Row():
                 lowvram_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="Low VRAM" if model_config_data["model_capabilties"]["lowvram_capable"] else "Low VRAM N/A", value="Enabled" if model_config_data["settings"]["lowvram_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["lowvram_capable"])
                 deepspeed_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="DeepSpeed Activate" if model_config_data["model_capabilties"]["deepspeed_capable"] else "DeepSpeed N/A", value="Enabled" if model_config_data["settings"]["deepspeed_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["deepspeed_capable"])
+                streaming_enabled_gr = gr.Radio(choices={"Enabled": "true", "Disabled": "false"}, label="Streaming" if model_config_data["model_capabilties"]["streaming_capable"] else "Streaming N/A", value="Enabled" if model_config_data["settings"]["streaming_enabled"] else "Disabled", interactive=model_config_data["model_capabilties"]["streaming_capable"])
                 temperature_set_gr = gr.Slider(value=float(model_config_data["settings"]["temperature_set"]), minimum=0, maximum=1, step=0.05, label="Temperature" if model_config_data["model_capabilties"]["temperature_capable"] else "Temperature N/A", interactive=model_config_data["model_capabilties"]["temperature_capable"])
                 repetitionpenalty_set_gr = gr.Slider(value=float(model_config_data["settings"]["repetitionpenalty_set"]), minimum=1, maximum=20, step=1, label="Repetition Penalty" if model_config_data["model_capabilties"]["repetitionpenalty_capable"] else "Repetition N/A", interactive=model_config_data["model_capabilties"]["repetitionpenalty_capable"])
                 pitch_set_gr = gr.Slider(value=float(model_config_data["settings"]["pitch_set"]), minimum=-10, maximum=10, step=1, label="Pitch" if model_config_data["model_capabilties"]["pitch_capable"] else "Pitch N/A", interactive=model_config_data["model_capabilties"]["pitch_capable"])
@@ -132,7 +134,7 @@ def parler_model_alltalk_settings(model_config_data):
             with gr.Row():
                 submit_button = gr.Button("Update Settings")
                 output_message = gr.Textbox(label="Output Message", interactive=False, show_label=False)
-            submit_button.click(parler_model_update_settings, inputs=[def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr, alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr], outputs=output_message)
+            submit_button.click(parler_model_update_settings, inputs=[def_character_voice_gr, def_narrator_voice_gr, lowvram_enabled_gr, deepspeed_enabled_gr, streaming_enabled_gr, temperature_set_gr, repetitionpenalty_set_gr, pitch_set_gr, generationspeed_set_gr, alloy_gr, echo_gr, fable_gr, nova_gr, onyx_gr, shimmer_gr], outputs=output_message)
             with gr.Accordion("HELP - 🔊 Understanding TTS Engine Default Settings Page", open=False):
                 with gr.Row():
                     gr.Markdown(AllTalkHelpContent.DEFAULT_SETTINGS, elem_classes="custom-markdown")                               

system/tts_engines/piper/help_content.py

@@ -237,6 +237,11 @@ class AllTalkHelpContent:
         - Accelerates TTS generation using optimized inference
         - Only available for engines and models that support DeepSpeed
         - Requires NVIDIA GPU with CUDA support
+
+    - **Stream Response Capability**
+        - Enables real-time streaming of generated speech output
+        - Reduces latency for faster feedback during synthesis
+        - Only available for engines and models that support Streaming
 
     - **Temperature Control**
         - Adjusts the variability in speech generation