saattrupdan commited on
Commit
4e86063
·
1 Parent(s): 8d98748

feat: Change to Whisper model

Browse files
Files changed (1) hide show
  1. app.py +21 -14
app.py CHANGED
@@ -23,18 +23,14 @@ load_dotenv()
23
 
24
  warnings.filterwarnings("ignore", category=FutureWarning)
25
 
26
- icon = """
27
- <svg xmlns="http://www.w3.org/2000/svg" width="14px" viewBox="0 0 24 24" fill="none"
28
- stroke="currentColor" stroke-width="2" stroke-linecap="round"
29
- stroke-linejoin="round" style="display: inline;">
30
- <path d="M21 15v4a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2v-4"/>
31
- <polyline points="17 8 12 3 7 8"/>
32
- <line x1="12" y1="3" x2="12" y2="15"/>
33
- </svg>
34
- """
35
  TITLE = "Røst Speech-to-text Demo"
 
36
  EMAIL_SUBJECT = "Røst tale-til-tekst demo".replace(" ", "+")
37
- EMAIL_BODY = """Hej,
 
38
 
39
  Jeg har lige prøvet jeres Røst tale-til-tekst demo, og jeg er imponeret!
40
 
@@ -43,15 +39,26 @@ Jeg kunne godt tænke mig at høre mere om jeres talegenkendelsesløsninger.
43
  Min use case er [indsæt use case her].
44
 
45
  Venlig hilsen,
46
- [dit navn]""".replace(" ", "+").replace("\n", "%0D")
 
 
 
 
 
 
 
 
 
 
 
47
  DESCRIPTION = f"""
48
  This is a demo of the Danish speech recognition model
49
- [Røst](https://huggingface.co/CoRal-project/roest-wav2vec2-315m-v2).
50
 
51
  Press "Record" to record your
52
  own voice. When you're done you can press "Stop" to stop recording and "Submit" to
53
  send the audio to the model for transcription. You can also upload an audio file by
54
- pressing the {icon} button.
55
 
56
  _If you like what you see and are interested in integrating speech-to-text solutions
57
  into your products, feel free to
@@ -62,7 +69,7 @@ logger.info("Loading the ASR model...")
62
  device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
63
  transcriber = pipeline(
64
  task="automatic-speech-recognition",
65
- model="CoRal-project/roest-wav2vec2-315m-v2",
66
  device=device,
67
  )
68
 
 
23
 
24
  warnings.filterwarnings("ignore", category=FutureWarning)
25
 
26
+
27
+ MODEL_ID = "alexandrainst/roest-whisper-1.5b-all-datasets"
28
+
 
 
 
 
 
 
29
  TITLE = "Røst Speech-to-text Demo"
30
+
31
  EMAIL_SUBJECT = "Røst tale-til-tekst demo".replace(" ", "+")
32
+ EMAIL_BODY = """
33
+ Hej,
34
 
35
  Jeg har lige prøvet jeres Røst tale-til-tekst demo, og jeg er imponeret!
36
 
 
39
  Min use case er [indsæt use case her].
40
 
41
  Venlig hilsen,
42
+ [dit navn]
43
+ """.strip().replace(" ", "+").replace("\n", "%0D")
44
+
45
+ ICON = """
46
+ <svg xmlns="http://www.w3.org/2000/svg" width="14px" viewBox="0 0 24 24" fill="none"
47
+ stroke="currentColor" stroke-width="2" stroke-linecap="round"
48
+ stroke-linejoin="round" style="display: inline;">
49
+ <path d="M21 15v4a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2v-4"/>
50
+ <polyline points="17 8 12 3 7 8"/>
51
+ <line x1="12" y1="3" x2="12" y2="15"/>
52
+ </svg>
53
+ """
54
  DESCRIPTION = f"""
55
  This is a demo of the Danish speech recognition model
56
+ [Røst](https://huggingface.co/{MODEL_ID}).
57
 
58
  Press "Record" to record your
59
  own voice. When you're done you can press "Stop" to stop recording and "Submit" to
60
  send the audio to the model for transcription. You can also upload an audio file by
61
+ pressing the {ICON} button.
62
 
63
  _If you like what you see and are interested in integrating speech-to-text solutions
64
  into your products, feel free to
 
69
  device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
70
  transcriber = pipeline(
71
  task="automatic-speech-recognition",
72
+ model=MODEL_ID,
73
  device=device,
74
  )
75