Merge branch 'robotic_voice'

README.md

@@ -22,6 +22,7 @@ preload_from_hub:
   - Pendrokar/xvapitch_nvidia
   - Pendrokar/TorchMoji
   - Pendrokar/xvasynth_lojban
+  - Pendrokar/xvasynth_cabal
 license: gpl-3.0
 short_description: CPU powered, low RTF, emotional, multilingual TTS
 ---

app.py

@@ -9,17 +9,29 @@ import resources.app.no_server as xvaserver
 
 from gr_client import BlocksDemo
 
-# model
+# NVIDIA NeMo models
 hf_model_name = "Pendrokar/xvapitch_nvidia"
 model_repo = HfApi()
 commits = model_repo.list_repo_commits(repo_id=hf_model_name)
 latest_commit_sha = commits[0].commit_id
 hf_cache_models_path = f'/home/user/.cache/huggingface/hub/models--Pendrokar--xvapitch_nvidia/snapshots/{latest_commit_sha}/'
+models_path = hf_cache_models_path
 
+# Lojban model
 commits = model_repo.list_repo_commits(repo_id='Pendrokar/xvasynth_lojban')
 latest_commit_sha = commits[0].commit_id
 hf_cache_lojban_models_path = f'/home/user/.cache/huggingface/hub/models--Pendrokar--xvasynth_lojban/snapshots/{latest_commit_sha}/'
-models_path = hf_cache_models_path
+
+# Robotic model
+hf_cache_robotic_models_path = ''
+try:
+	commits = model_repo.list_repo_commits(repo_id='Pendrokar/xvasynth_cabal', token=os.getenv('HF_TOKEN'))
+	latest_commit_sha = commits[0].commit_id
+	hf_cache_robotic_models_path = f'/home/user/.cache/huggingface/hub/models--Pendrokar--xvasynth_cabal/snapshots/{latest_commit_sha}/'
+except:
+	print('Robotic voice not loaded!')
+	pass
+
 
 current_voice_model = None
 current_voice_type = None
@@ -33,7 +45,10 @@ def load_model(voice_model_name):
 		model_path =  hf_cache_lojban_models_path + voice_model_name
 		model_type = 'FastPitch1.1'
 	else:
-		model_path =  models_path + voice_model_name
+		if voice_model_name == 'cnc_cabal':
+			model_path =  hf_cache_robotic_models_path + voice_model_name
+		else:
+			model_path =  models_path + voice_model_name
 		model_type = 'xVAPitch'
 
 	language = 'en' # seems to have no effect if generated text is from a different language
@@ -199,5 +214,5 @@ class LocalBlocksDemo(BlocksDemo):
 
 if __name__ == "__main__":
 	print('running custom Gradio interface')
-	demo = LocalBlocksDemo(models_path, hf_cache_lojban_models_path)
+	demo = LocalBlocksDemo(models_path, hf_cache_lojban_models_path, hf_cache_robotic_models_path)
 	demo.block.launch()

gr_client.py

@@ -316,9 +316,17 @@ _DESCRIPTION = '''
 
 
 class BlocksDemo:
-	def __init__(self, models_path, lojban_models_path):
+	def __init__(self, models_path, lojban_models_path, robotic_models_path):
+		# global voice_radio_init
 		self.models_path = models_path
 		self.lojban_models_path = lojban_models_path
+		self.robotic_models_path = robotic_models_path
+
+		if robotic_models_path != '':
+			# insert robotic voice as the third option
+			voice_models.append(("🤖 Robot", "cnc_cabal"))
+			voice_radio_init['choices'].insert(2, ("🤖 Robot", "cnc_cabal"))
+
 		self.block = self.create_interface()
 
 	def create_interface(self):
@@ -605,6 +613,8 @@ class BlocksDemo:
 
 		if voice_id == 'x_selpahi':
 			sample_path = self.lojban_models_path
+		if voice_id == 'cnc_cabal':
+			sample_path = self.robotic_models_path
 		else:
 			sample_path = self.models_path
 
@@ -620,5 +630,5 @@ if __name__ == "__main__":
 	print('running Gradio interface')
 	client = Client("Pendrokar/xVASynth")
 
-	demo = BlocksDemo('', '')
+	demo = BlocksDemo('', '', '')
 	demo.block.launch()