cati
commited on
Commit
•
90f256f
1
Parent(s):
69d94dc
..
Browse files- .#ctcalign.py +1 -0
- app.py +8 -1
- ctcalign.py +12 -1
.#ctcalign.py
ADDED
@@ -0,0 +1 @@
|
|
|
app.py
CHANGED
@@ -61,10 +61,17 @@ All phoneme durations are measured automatically with no human correction. The p
|
|
61 |
gr.Markdown(
|
62 |
"""
|
63 |
# Forced alignment demo with CTC decoding
|
64 |
-
Choose a language to record or upload a sentence with text. Generate word
|
65 |
"""
|
66 |
)
|
67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
|
69 |
|
70 |
|
|
|
61 |
gr.Markdown(
|
62 |
"""
|
63 |
# Forced alignment demo with CTC decoding
|
64 |
+
Choose a language to record or upload a sentence with corresponding text. Generate word and letter time-alignments from the language's wav2vec-2.0 model, with output in MFA (Montreal Forced Aligner)-compatible format.
|
65 |
"""
|
66 |
)
|
67 |
|
68 |
+
audio_file = gr.Audio(type="filepath")
|
69 |
+
transcript_boxx = gr.Textbox(label="Transcript",placeholder="Type or paste the transcript here. Capitalisation and punctuation, if any, will be ignored.")
|
70 |
+
alangmenu = gr.Radio(["Icelandic", "Faroese", "Norwegian"],value="Icelandic")
|
71 |
+
|
72 |
+
al_btn = gr.Button(value="Run forced alignment")
|
73 |
+
|
74 |
+
output_box = gr.Textbox(label="Forced alignment output")
|
75 |
|
76 |
|
77 |
|
ctcalign.py
CHANGED
@@ -11,7 +11,7 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
11 |
torch.random.manual_seed(0)
|
12 |
|
13 |
# info: https://huggingface.co/carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h/blob/main/vocab.json
|
14 |
-
MODEL_PATH="/
|
15 |
model_blank_token = '[PAD]' # important to know for CTC decoding
|
16 |
model_word_separator = '|'
|
17 |
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}
|
@@ -25,6 +25,17 @@ blank_id = labels_dict[model_blank_token]
|
|
25 |
|
26 |
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
#------------------------------------------
|
29 |
# forced alignment with ctc decoder
|
30 |
# originally based on implementation of
|
|
|
11 |
torch.random.manual_seed(0)
|
12 |
|
13 |
# info: https://huggingface.co/carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h/blob/main/vocab.json
|
14 |
+
MODEL_PATH="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-icelandic-ep10-1000h"
|
15 |
model_blank_token = '[PAD]' # important to know for CTC decoding
|
16 |
model_word_separator = '|'
|
17 |
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}
|
|
|
25 |
|
26 |
|
27 |
|
28 |
+
|
29 |
+
|
30 |
+
|
31 |
+
|
32 |
+
|
33 |
+
|
34 |
+
|
35 |
+
|
36 |
+
|
37 |
+
|
38 |
+
|
39 |
#------------------------------------------
|
40 |
# forced alignment with ctc decoder
|
41 |
# originally based on implementation of
|