app

app.py

@@ -0,0 +1,78 @@
+import gradio as gr
+import subprocess
+import scripts.runSQ
+
+
+#https://huggingface.co/spaces/clr/prosalign/blob/main/app.py
+
+
+def setup():
+    r0 = subprocess.run(["pwd"], capture_output=True, text=True)
+    print('PWD::', r0.stdout)
+    r1 = subprocess.run(["wget", "https://github.com/google/REAPER/archive/refs/heads/master.zip"], capture_output=True, text=True)
+    print(r1.stdout)
+    subprocess.run(["unzip", "./master.zip"])
+    subprocess.run(["mv", "REAPER-master", "REAPER"])
+    subprocess.run(["rm", "./master.zip"])
+    os.chdir('./REAPER')
+    subprocess.run(["mkdir", "build"])
+    os.chdir('./build')
+    r2 = subprocess.run(["cmake", ".."], capture_output=True, text=True)
+    print(r2.stdout)
+    r3 = subprocess.run(["make"], capture_output=True, text=True)
+    print(r3.stdout)
+    
+    os.chdir('../..')
+    r9 = subprocess.run(["ls", "-la"], capture_output=True, text=True)
+    print('LS::', r9.stdout)
+
+                        
+print('about to setup')
+setup()
+
+
+def f1(voices, sent):
+    one_tts = scripts.runSQ.run(sentence,voices)
+    return (one_tts)
+
+
+
+
+
+bl = gr.Blocks()
+with bl:
+
+
+    temp_sentences = ['Litlaus græn hugmynd?','Var það ekki nóg?', 'Ef svo er hvað heita þau þá?','Eru maríuhænur á Íslandi?']
+    
+    voices = ['Alfur','Dora']
+    # currently i only get json speech marks for those two. 
+    # supposedly they also provided for Karl, Dora, but i dont even get their wavs
+    # i get everyone elses wavs tho
+    
+    with gr.Row():
+        with gr.column(scale=4):
+            voiceselect = gr.checkboxgroup(voices, label="TTS voice",value='Alfur')
+            temp_sentmenu = gr.dropdown(temp_sentences, label="Sentence")
+        with gr.column(scale=1:)
+            temp_button = gr.Button(value="A button")
+    
+    
+    tts_output = gr.Audio(interactive=False)
+    
+    
+    
+    tmp_button.click(f1,[voiceselect,temp_sentmenu],[tts_output])
+    
+    
+if __name__ == "__main__":
+    bl.launch()    
+    
+    
+    
+    
+    
+    
+    
+    
+    

scripts/runSQ.py

@@ -8,30 +8,41 @@ from tapi import tiro
 #  report how many, or if 0.
 
 
-def run():
-    sentence = 'hvaða sjúkdómar geta fylgt óbeinum reykingum'
-    voices = ['Alfur','Dilja','Karl', 'Dora']
+def run(sentence, voices):
+    #sentence = 'hvaða sjúkdómar geta fylgt óbeinum reykingum'
+    #voices = ['Alfur','Dilja','Karl', 'Dora']
     # On tts.tiro.is speech marks are only available 
     # for the voices: Alfur, Dilja, Karl and Dora.
 
-    corpus_meta = 'human_data/SQL1adult_metadata.tsv'
-    speech_dir = 'human_data/audio/squeries/'
-    speech_aligns = 'human_data/aligns/squeries/'
-    speech_f0 = 'human_data/f0/squeries/'
+    corpus_meta = '../human_data/SQL1adult_metadata.tsv'
+    speech_dir = '../human_data/audio/squeries/'
+    speech_aligns = '../human_data/aligns/squeries/'
+    speech_f0 = '../human_data/f0/squeries/'
     align_model_path ="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h"
 
-    tts_dir = 'tts_data/'
-
+    tts_dir = '../tts_data/'
+    
+    
+    norm_sentencd = snorm(sentence)
 
-    meta = get_recordings(sentence, corpus_meta)
+    meta = get_recordings(norm_sentence, corpus_meta)
     if meta:
         align_human(meta,speech_aligns,speech_dir,align_model_path)
         f0_human(meta, speech_f0, speech_dir, 'TODO path to reaper')
     if voices:
-        get_tts(sentence,voices,tts_dir)
+        temp_a_sample = get_tts(sentence,voices,tts_dir)
         f0_tts(sentence, voices, tts_dir, 'TODO path to reaper')
+        
+    return temp_a_sample
 
 
+def snorm(s):
+    s = ''.join([c.lower() for c in s if not unicodedata.category(c).startswith("P") ])
+    while '  ' in s:
+        s = s.replace('  ', ' ')
+    return s
+    
+
 # find all the recordings of a given sentence
 # listed in the corpus metadata.
 # sentence should be provided lowercase without punctuation
@@ -122,12 +133,16 @@ def get_tts(sentence,voices,ttsdir):
     dpath = sentence.replace(' ','_')[:65]
     
     no_voice = []
+    
+    temp_sample_path = ''
 
     for v in voices:
         wpath = f'{ttsdir}{dpath}/{v}.wav'
         jpath = f'{ttsdir}{dpath}/{v}.json'
         if not (os.path.exists(wpath) and os.path.exists(jpath)):
             no_voice.append(v)
+        if not temp_sample_path:
+            temp_sample_path = wpath
             
     if no_voice:
         print(f'Need to generate TTS for {len(no_voice)} voices')
@@ -138,6 +153,8 @@ def get_tts(sentence,voices,ttsdir):
         
     else:
         print('TTS for all voices existed')
+        
+    return temp_sample_path