tune voice for audiobook

.gitattributes

@@ -46,4 +46,4 @@ uc_spk_Landscape2Soundscape_Masterpieces_pics/10_Boecklin_967648_NG2-80_001_rsz.
 uc_spk_Landscape2Soundscape_Masterpieces_pics/03_Schinkel_WS200-002.jpg filter=lfs diff=lfs merge=lfs -text
 uc_spk_Landscape2Soundscape_Masterpieces_pics/11_Liebermann_NG4-94_001.jpg filter=lfs diff=lfs merge=lfs -text
 uc_spk_Landscape2Soundscape_Masterpieces_pics/12_Slevogt_AII1022_001.jpg filter=lfs diff=lfs merge=lfs -text
-audiobook_TTS.docx filter=lfs diff=lfs merge=lfs -text
+assets/audiobook_TTS.docx filter=lfs diff=lfs merge=lfs -text

README.md

@@ -26,10 +26,10 @@ Beta version of [SHIFT](https://shift-europe.eu/) TTS tool with [AudioGen sounds
   - [Analysis of emotion of SHIFT TTS](https://huggingface.co/dkounadis/artificial-styletts2/discussions/2)
   - [Listen Also foreign languages](https://huggingface.co/dkounadis/artificial-styletts2/discussions/4) synthesized via [MMS TTS](https://huggingface.co/facebook/mms-tts)
 
-## Available Voices
+## Listen to Available Voices!
 
 
-<a href="https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#67854dcbd3e6beb1a78f7f20">Native English!</a> / <a href="https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#6783e3b00e7d90facec060c6">Non-native English accents!</a> / <a href="https://huggingface.co/dkounadis/artificial-styletts2/blob/main/Utils/all_langs.csv">Foreign languages</a>
+<a href="https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#67854dcbd3e6beb1a78f7f20">Native English</a> / <a href="https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#6783e3b00e7d90facec060c6">Non-native English: Accents</a> / <a href="https://huggingface.co/dkounadis/artificial-styletts2/blob/main/Utils/all_langs.csv">Foreign languages</a>
 
 ##
 
@@ -62,16 +62,13 @@ Flask `tmux-session`
 CUDA_DEVICE_ORDER=PCI_BUS_ID HF_HOME=./hf_home CUDA_VISIBLE_DEVICES=0 python api.py
 ```
 
-Inference via Flask
-
-Following examples need `api.py` to be running. [Set this IP](https://huggingface.co/dkounadis/artificial-styletts2/blob/main/tts.py#L85) to the IP shown when starting `api.py`.
-
-## Examples For Cultural Heritage Partners - SHIFT
+Inference via Flask. Following examples need `api.py` to be running. [Set this IP](https://huggingface.co/dkounadis/artificial-styletts2/blob/main/tts.py#L85) to the IP shown when starting `api.py`.
 
 </details>
 
+# SHIFT Demos
 
-## Landscape 2 Soundscape
+## Landscape 2 Soundscapes
 
 The following needs `api.py` to be already running on a tmux session. 
 

api.py

@@ -170,7 +170,7 @@ def tts_multi_sentence(precomputed_style_vector=None,
                         precomputed_style_vector,
                                     alpha=0.3,
                                     beta=0.7,
-                                    diffusion_steps=7,
+                                    diffusion_steps=diffusion_steps,
                                     embedding_scale=1))
         x = np.concatenate(x)
                     
@@ -286,8 +286,10 @@ def serve_wav():
     
     # Native Eng
 
-    if precomputed_style_vector is None:  # TODO 7 diffusion steps 
+    if precomputed_style_vector is None:
+        
         if 'en_US' in args.voice or 'en_UK' in args.voice:
+            diffusion_steps = 7  # native
             _dir = '/' if args.affective else '_v2/'
             precomputed_style_vector = msinference.compute_style(
                 'assets/wavs/style_vector' + _dir + args.voice.replace(
@@ -298,11 +300,12 @@ def serve_wav():
         
         # Non-Native Eng
         
-        elif '_' in  args.voice:  # TODO 5 diffusion steps
+        elif '_' in  args.voice:
             precomputed_style_vector = msinference.compute_style('assets/wavs/mimic3_foreign_4x/' + args.voice.replace(
                                                                  '/', '_').replace('#', '_').replace(
                                                                     'cmu-arctic', 'cmu_arctic').replace(
                                                                         '_low', '') + '.wav')
+            diffusion_steps = 5  # non-native
         
         # Foreign Lang - MMS/TTS 
         else:

audiobook.py

@@ -19,13 +19,14 @@ Path(ROOT_DIR).mkdir(parents=True,
                      exist_ok=True)
 voices = [
     # 'en_US/vctk_low#p228',  # https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#67854dcbd3e6beb1a78f7f20
-    'af_ZA_google-nwu_0184',  # https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#6783e3b00e7d90facec060c6
-    # 'en_US/vctk_low#p326',
+    # 'af_ZA_google-nwu_0184',  # https://huggingface.co/dkounadis/artificial-styletts2/discussions/1#6783e3b00e7d90facec060c6
+    'en_US/vctk_low#p326',   # Native voice
+#    'jv_ID_google-gmu_06207',
     ]  # select any voice from - https://audeering.github.io/shift/
 
 #urllib.request.urlretrieve("https://github.com/audeering/shift/raw/refs/heads/main/assets/INCLUSION_IN_MUSEUMS_audiobook.docx", "audiobook_TTS.docx")
 
-d = docx.Document('audiobook_TTS.docx')  # slightly changed from the original .docx to be audible as by adding extra 'by them from this of etc.'
+d = docx.Document('assets/audiobook_TTS.docx')  # slightly changed from the original .docx to be audible as by adding extra 'by them from this of etc.'
 
 last_paragraph_was_silence = False  # to know to add silence only once after only at the 1st empty paragraph we detect
 

msinference.py

@@ -173,7 +173,7 @@ def inference(text,
               ref_s,
               alpha = 0.3,
               beta = 0.7,
-              diffusion_steps=5,   # diffusion_steps=5 sounds more pleasant than 7
+              diffusion_steps=7, # 7 if voice is native English else 5 for non-native
               embedding_scale=1,
               use_gruut=False):
     text = text.strip()