..

.#ctcalign.py

@@ -1 +0,0 @@
-[email protected]

app.py

@@ -65,13 +65,17 @@ All phoneme durations are measured automatically with no human correction. The p
  """
  )
 
- audio_file = gr.Audio(type="filepath")
- transcript_boxx = gr.Textbox(label="Transcript",placeholder="Type or paste the transcript here. Capitalisation and punctuation, if any, will be ignored.")
- alangmenu = gr.Radio(["Icelandic", "Faroese", "Norwegian"],value="Icelandic")
+ with gr.Row():
+ with gr.Column():
+ transcript_boxx = gr.Textbox(label="Transcript",placeholder="Type or paste the transcript here. Capitalisation and punctuation, if any, will be ignored.")
+ alangmenu = gr.Radio(["Icelandic", "Faroese", "Norwegian"],value="Icelandic")
 
- al_btn = gr.Button(value="Run forced alignment")
+ audio_file = gr.Audio(type="filepath")
 
- output_box = gr.Textbox(label="Forced alignment output")
+ al_btn = gr.Button(value="Run forced alignment")
+
+ with gr.Column():
+ output_box = gr.Textbox(label="Forced alignment output")
 
 
 

ctcalign.py

@@ -11,23 +11,43 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 torch.random.manual_seed(0)
 
 # info: https://huggingface.co/carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h/blob/main/vocab.json
-MODEL_PATH="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h"
-model_blank_token = '[PAD]' # important to know for CTC decoding
-model_word_separator = '|'
-labels_dict = {"f": 0, "a": 1, "é": 2, "t": 3, "o": 4, "n": 5, "e": 6, "y": 8, "k": 9, "j": 10, "u": 11, "d": 12, "w": 13, "l": 14, "ú": 15, "q": 16, "g": 17, "í": 18, "s": 19, "r": 20, "ý": 21, "i": 22, "z": 23, "m": 24, "h": 25, "ó": 26, "þ": 27, "æ": 28, "c": 29, "á": 30, "v": 31, "b": 32, "ð": 33, "x": 34, "ö": 35, "p": 36, "|": 7, "[UNK]": 37, "[PAD]": 38}
+is_MODEL_PATH="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h"
+is_model_blank_token = '[PAD]' # important to know for CTC decoding
+is_model_word_separator = '|'
+is_labels_dict = {"f": 0, "a": 1, "é": 2, "t": 3, "o": 4, "n": 5, "e": 6, "y": 8, "k": 9, "j": 10, "u": 11, "d": 12, "w": 13, "l": 14, "ú": 15, "q": 16, "g": 17, "í": 18, "s": 19, "r": 20, "ý": 21, "i": 22, "z": 23, "m": 24, "h": 25, "ó": 26, "þ": 27, "æ": 28, "c": 29, "á": 30, "v": 31, "b": 32, "ð": 33, "x": 34, "ö": 35, "p": 36, "|": 7, "[UNK]": 37, "[PAD]": 38}
 
-model = Wav2Vec2ForCTC.from_pretrained(MODEL_PATH).to(device)
-processor = Wav2Vec2Processor.from_pretrained(MODEL_PATH)
-inverse_dict = {v:k for k,v in labels_dict.items()}
-all_labels = tuple(labels_dict.keys())
-blank_id = labels_dict[model_blank_token]
+is_model = Wav2Vec2ForCTC.from_pretrained(is_MODEL_PATH).to(device)
+is_processor = Wav2Vec2Processor.from_pretrained(is_MODEL_PATH)
+is_inverse_dict = {v:k for k,v in is_labels_dict.items()}
+is_all_labels = tuple(is_labels_dict.keys())
+is_blank_id = is_labels_dict[is_model_blank_token]
 
 
 
+fo_MODEL_PATH="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-faroese-100h"
+fo_model_blank_token = '[PAD]' # important to know for CTC decoding
+fo_model_word_separator = '|'
+fo_labels_dict = {"w": 0, "i": 1, "6": 2, "s": 3, "_": 4, "k": 5, "l": 6, "ú": 7, "2": 8, "4": 9, "d": 10, "z": 11, "3": 12, "ð": 13, "t": 15, "ø": 16, "x": 17, "p": 18, "o": 19, "æ": 20, "n": 21, "f": 22, "á": 23, "5": 24, "g": 25, "ý": 26, "r": 27, "é": 28, "u": 29, "ü": 30, "y": 31, "í": 32, "h": 33, "q": 34, "b": 35, "e": 36, "v": 37, "-": 38, "c": 39, "j": 40, ".": 41, "ó": 42, "'": 43, "m": 44, "a": 45, "|": 14, "[UNK]": 46, "[PAD]": 47}
 
+fo_model = Wav2Vec2ForCTC.from_pretrained(fo_MODEL_PATH).to(device)
+fo_processor = Wav2Vec2Processor.from_pretrained(fo_MODEL_PATH)
+fo_inverse_dict = {v:k for k,v in fo_labels_dict.items()}
+fo_all_labels = tuple(fo_labels_dict.keys())
+fo_blank_id = fo_labels_dict[fo_model_blank_token]
 
 
 
+no_MODEL_PATH="NbAiLab/nb-wav2vec2-1b-bokmaal"
+no_model_blank_token = '[PAD]' # important to know for CTC decoding
+no_model_word_separator = '|'
+no_labels_dict = {"a": 1, "b": 2, "c": 3, "d": 4, "e": 5, "f": 6, "g": 7, "h": 8, "i": 9, "j": 10, "k": 11, "l": 12, "m": 13, "n": 14, "o": 15, "p": 16, "q": 17, "r": 18, "s": 19, "t": 20, "u": 21, "v": 22, "w": 23, "x": 24, "y": 25, "z": 26, "å": 27, "æ": 28, "ø": 29, "|": 0, "[UNK]": 30, "[PAD]": 31}
+
+no_model = Wav2Vec2ForCTC.from_pretrained(no_MODEL_PATH).to(device)
+no_processor = Wav2Vec2Processor.from_pretrained(no_MODEL_PATH)
+no_inverse_dict = {v:k for k,v in no_labels_dict.items()}
+no_all_labels = tuple(no_labels_dict.keys())
+no_blank_id = no_labels_dict[no_model_blank_token]
+