sanchit-gandhi commited on
Commit
e95e308
·
1 Parent(s): 2f42453

update examples

Browse files
Files changed (1) hide show
  1. app.py +7 -8
app.py CHANGED
@@ -10,8 +10,7 @@ from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed
10
 
11
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
12
 
13
- # TODO(SG): update to the latest checkpoint
14
- repo_id = "reach-vb/parler-tts-expresso-mistral-v0.1"
15
 
16
  model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
17
  tokenizer = AutoTokenizer.from_pretrained(repo_id)
@@ -21,18 +20,18 @@ SAMPLE_RATE = feature_extractor.sampling_rate
21
  SEED = 42
22
 
23
  default_text = "*Remember* - this is only the first iteration of the model! To improve the prosody and naturalness of the speech further, we're scaling up the amount of training data by a factor of *five times*."
24
- default_description = "Thomas speaks with emphasis at a moderate pace with high quality."
25
  examples = [
26
  [
27
- "Remember - this is only the first iteration of the model! To improve the prosody and naturalness of the speech further, we're scaling up the amount of training data by a factor of five times.",
28
- "Thomas speaks sadly at a very slow pace with high quality."
29
  ],
30
  [
31
- "Shhh! Did you know? You can reproduce this entire training recipe by following the steps outlined on the model card. It only takes one hour to train!",
32
- "Talia whispers quickly with high quality audio.",
33
  ],
34
  [
35
- "But that's no secret! The entire project is open-source first. We are releasing all datasets, training and inference code, so that you can use them yourself!",
36
  "Elisabeth speaks happily at a slightly slower than average pace with high quality audio.",
37
  ],
38
  [
 
10
 
11
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
12
 
13
+ repo_id = "parler-tts/parler-tts-mini-expresso"
 
14
 
15
  model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
16
  tokenizer = AutoTokenizer.from_pretrained(repo_id)
 
20
  SEED = 42
21
 
22
  default_text = "*Remember* - this is only the first iteration of the model! To improve the prosody and naturalness of the speech further, we're scaling up the amount of training data by a factor of *five times*."
23
+ default_description = "Thomas speaks with emphasis and excitement at a moderate pace with high quality."
24
  examples = [
25
  [
26
+ "Remember - this is only the first iteration of the model. To improve the prosody and naturalness of the speech further, we're scaling up the amount of training data by a factor of five times.",
27
+ "Thomas speaks in a sad tone at a very slow pace with high quality."
28
  ],
29
  [
30
+ "Did you know? <laugh> You can reproduce this entire training recipe by following the steps outlined on the model card!",
31
+ "Talia speaks quickly with laughter and high quality audio.",
32
  ],
33
  [
34
+ "But that's no secret! The entire project is open-source first, with all release artefacts on the Hub.",
35
  "Elisabeth speaks happily at a slightly slower than average pace with high quality audio.",
36
  ],
37
  [