add new prompts

sanjayws · Apr 17, 2023 · 2345898 · 2345898
1 parent d621ee3
commit 2345898
Show file tree

Hide file tree

Showing 157 changed files with 28 additions and 5 deletions.
diff --git a/README.md b/README.md
@@ -84,7 +84,7 @@ text_prompt = """
     I have a silky smooth voice, and today I will tell you about 
     the exercise regimen of the common sloth.
 """
-audio_array = generate_audio(text_prompt, history_prompt="speech_0")
+audio_array = generate_audio(text_prompt, history_prompt="en_speaker_1")
 ```
 
 [sloth.webm](https://user-images.githubusercontent.com/5068315/230684883-a344c619-a560-4ff5-8b99-b4463a34487b.webm)

diff --git a/bark/assets/prompts/speech_0.npz → bark/assets/prompts/announcer.npz b/bark/assets/prompts/speech_0.npz → bark/assets/prompts/announcer.npz
diff --git a/bark/assets/prompts/de_speaker_0.npz b/bark/assets/prompts/de_speaker_0.npz
diff --git a/bark/assets/prompts/de_speaker_1.npz b/bark/assets/prompts/de_speaker_1.npz
diff --git a/bark/assets/prompts/de_speaker_2.npz b/bark/assets/prompts/de_speaker_2.npz
diff --git a/bark/assets/prompts/de_speaker_3.npz b/bark/assets/prompts/de_speaker_3.npz
diff --git a/bark/assets/prompts/de_speaker_4.npz b/bark/assets/prompts/de_speaker_4.npz
diff --git a/bark/assets/prompts/de_speaker_5.npz b/bark/assets/prompts/de_speaker_5.npz
diff --git a/bark/assets/prompts/de_speaker_6.npz b/bark/assets/prompts/de_speaker_6.npz
diff --git a/bark/assets/prompts/de_speaker_7.npz b/bark/assets/prompts/de_speaker_7.npz
diff --git a/bark/assets/prompts/de_speaker_8.npz b/bark/assets/prompts/de_speaker_8.npz
diff --git a/bark/assets/prompts/de_speaker_9.npz b/bark/assets/prompts/de_speaker_9.npz
diff --git a/bark/assets/prompts/en_speaker_0.npz b/bark/assets/prompts/en_speaker_0.npz
diff --git a/bark/assets/prompts/en_speaker_1.npz b/bark/assets/prompts/en_speaker_1.npz
diff --git a/bark/assets/prompts/en_speaker_2.npz b/bark/assets/prompts/en_speaker_2.npz
diff --git a/bark/assets/prompts/en_speaker_3.npz b/bark/assets/prompts/en_speaker_3.npz
diff --git a/bark/assets/prompts/en_speaker_4.npz b/bark/assets/prompts/en_speaker_4.npz
diff --git a/bark/assets/prompts/en_speaker_5.npz b/bark/assets/prompts/en_speaker_5.npz
diff --git a/bark/assets/prompts/en_speaker_6.npz b/bark/assets/prompts/en_speaker_6.npz
diff --git a/bark/assets/prompts/en_speaker_7.npz b/bark/assets/prompts/en_speaker_7.npz
diff --git a/bark/assets/prompts/en_speaker_8.npz b/bark/assets/prompts/en_speaker_8.npz
diff --git a/bark/assets/prompts/en_speaker_9.npz b/bark/assets/prompts/en_speaker_9.npz
diff --git a/bark/assets/prompts/es_speaker_0.npz b/bark/assets/prompts/es_speaker_0.npz
diff --git a/bark/assets/prompts/es_speaker_1.npz b/bark/assets/prompts/es_speaker_1.npz
diff --git a/bark/assets/prompts/es_speaker_2.npz b/bark/assets/prompts/es_speaker_2.npz
diff --git a/bark/assets/prompts/es_speaker_3.npz b/bark/assets/prompts/es_speaker_3.npz
diff --git a/bark/assets/prompts/es_speaker_4.npz b/bark/assets/prompts/es_speaker_4.npz
diff --git a/bark/assets/prompts/es_speaker_5.npz b/bark/assets/prompts/es_speaker_5.npz
diff --git a/bark/assets/prompts/es_speaker_6.npz b/bark/assets/prompts/es_speaker_6.npz
diff --git a/bark/assets/prompts/es_speaker_7.npz b/bark/assets/prompts/es_speaker_7.npz
diff --git a/bark/assets/prompts/es_speaker_8.npz b/bark/assets/prompts/es_speaker_8.npz
diff --git a/bark/assets/prompts/es_speaker_9.npz b/bark/assets/prompts/es_speaker_9.npz
diff --git a/bark/assets/prompts/fr_speaker_0.npz b/bark/assets/prompts/fr_speaker_0.npz
diff --git a/bark/assets/prompts/fr_speaker_1.npz b/bark/assets/prompts/fr_speaker_1.npz
diff --git a/bark/assets/prompts/fr_speaker_2.npz b/bark/assets/prompts/fr_speaker_2.npz
diff --git a/bark/assets/prompts/fr_speaker_3.npz b/bark/assets/prompts/fr_speaker_3.npz
diff --git a/bark/assets/prompts/fr_speaker_4.npz b/bark/assets/prompts/fr_speaker_4.npz
diff --git a/bark/assets/prompts/fr_speaker_5.npz b/bark/assets/prompts/fr_speaker_5.npz
diff --git a/bark/assets/prompts/fr_speaker_6.npz b/bark/assets/prompts/fr_speaker_6.npz
diff --git a/bark/assets/prompts/fr_speaker_7.npz b/bark/assets/prompts/fr_speaker_7.npz
diff --git a/bark/assets/prompts/fr_speaker_8.npz b/bark/assets/prompts/fr_speaker_8.npz
diff --git a/bark/assets/prompts/fr_speaker_9.npz b/bark/assets/prompts/fr_speaker_9.npz
diff --git a/bark/assets/prompts/hi_speaker_0.npz b/bark/assets/prompts/hi_speaker_0.npz
diff --git a/bark/assets/prompts/hi_speaker_1.npz b/bark/assets/prompts/hi_speaker_1.npz
diff --git a/bark/assets/prompts/hi_speaker_2.npz b/bark/assets/prompts/hi_speaker_2.npz
diff --git a/bark/assets/prompts/hi_speaker_3.npz b/bark/assets/prompts/hi_speaker_3.npz
diff --git a/bark/assets/prompts/hi_speaker_4.npz b/bark/assets/prompts/hi_speaker_4.npz
diff --git a/bark/assets/prompts/hi_speaker_5.npz b/bark/assets/prompts/hi_speaker_5.npz
diff --git a/bark/assets/prompts/hi_speaker_6.npz b/bark/assets/prompts/hi_speaker_6.npz
diff --git a/bark/assets/prompts/hi_speaker_7.npz b/bark/assets/prompts/hi_speaker_7.npz
diff --git a/bark/assets/prompts/hi_speaker_8.npz b/bark/assets/prompts/hi_speaker_8.npz
diff --git a/bark/assets/prompts/hi_speaker_9.npz b/bark/assets/prompts/hi_speaker_9.npz
diff --git a/bark/assets/prompts/it_speaker_0.npz b/bark/assets/prompts/it_speaker_0.npz
diff --git a/bark/assets/prompts/it_speaker_1.npz b/bark/assets/prompts/it_speaker_1.npz
diff --git a/bark/assets/prompts/it_speaker_2.npz b/bark/assets/prompts/it_speaker_2.npz
diff --git a/bark/assets/prompts/it_speaker_3.npz b/bark/assets/prompts/it_speaker_3.npz
diff --git a/bark/assets/prompts/it_speaker_4.npz b/bark/assets/prompts/it_speaker_4.npz
diff --git a/bark/assets/prompts/it_speaker_5.npz b/bark/assets/prompts/it_speaker_5.npz
diff --git a/bark/assets/prompts/it_speaker_6.npz b/bark/assets/prompts/it_speaker_6.npz
diff --git a/bark/assets/prompts/it_speaker_7.npz b/bark/assets/prompts/it_speaker_7.npz
diff --git a/bark/assets/prompts/it_speaker_8.npz b/bark/assets/prompts/it_speaker_8.npz
diff --git a/bark/assets/prompts/it_speaker_9.npz b/bark/assets/prompts/it_speaker_9.npz
diff --git a/bark/assets/prompts/ja_speaker_0.npz b/bark/assets/prompts/ja_speaker_0.npz
diff --git a/bark/assets/prompts/ja_speaker_1.npz b/bark/assets/prompts/ja_speaker_1.npz
diff --git a/bark/assets/prompts/ja_speaker_2.npz b/bark/assets/prompts/ja_speaker_2.npz
diff --git a/bark/assets/prompts/ja_speaker_3.npz b/bark/assets/prompts/ja_speaker_3.npz
diff --git a/bark/assets/prompts/ja_speaker_4.npz b/bark/assets/prompts/ja_speaker_4.npz
diff --git a/bark/assets/prompts/ja_speaker_5.npz b/bark/assets/prompts/ja_speaker_5.npz
diff --git a/bark/assets/prompts/ja_speaker_6.npz b/bark/assets/prompts/ja_speaker_6.npz
diff --git a/bark/assets/prompts/ja_speaker_7.npz b/bark/assets/prompts/ja_speaker_7.npz
diff --git a/bark/assets/prompts/ja_speaker_8.npz b/bark/assets/prompts/ja_speaker_8.npz
diff --git a/bark/assets/prompts/ja_speaker_9.npz b/bark/assets/prompts/ja_speaker_9.npz
diff --git a/bark/assets/prompts/ko_speaker_0.npz b/bark/assets/prompts/ko_speaker_0.npz
diff --git a/bark/assets/prompts/ko_speaker_1.npz b/bark/assets/prompts/ko_speaker_1.npz
diff --git a/bark/assets/prompts/ko_speaker_2.npz b/bark/assets/prompts/ko_speaker_2.npz
diff --git a/bark/assets/prompts/ko_speaker_3.npz b/bark/assets/prompts/ko_speaker_3.npz
diff --git a/bark/assets/prompts/ko_speaker_4.npz b/bark/assets/prompts/ko_speaker_4.npz
diff --git a/bark/assets/prompts/ko_speaker_5.npz b/bark/assets/prompts/ko_speaker_5.npz
diff --git a/bark/assets/prompts/ko_speaker_6.npz b/bark/assets/prompts/ko_speaker_6.npz
diff --git a/bark/assets/prompts/ko_speaker_7.npz b/bark/assets/prompts/ko_speaker_7.npz
diff --git a/bark/assets/prompts/ko_speaker_8.npz b/bark/assets/prompts/ko_speaker_8.npz
diff --git a/bark/assets/prompts/ko_speaker_9.npz b/bark/assets/prompts/ko_speaker_9.npz
diff --git a/bark/assets/prompts/music_0.npz b/bark/assets/prompts/music_0.npz
diff --git a/bark/assets/prompts/music_1.npz b/bark/assets/prompts/music_1.npz
diff --git a/bark/assets/prompts/music_2.npz b/bark/assets/prompts/music_2.npz
diff --git a/bark/assets/prompts/music_3.npz b/bark/assets/prompts/music_3.npz
diff --git a/bark/assets/prompts/music_4.npz b/bark/assets/prompts/music_4.npz
diff --git a/bark/assets/prompts/music_5.npz b/bark/assets/prompts/music_5.npz
diff --git a/bark/assets/prompts/pl_speaker_0.npz b/bark/assets/prompts/pl_speaker_0.npz
diff --git a/bark/assets/prompts/pl_speaker_1.npz b/bark/assets/prompts/pl_speaker_1.npz
diff --git a/bark/assets/prompts/pl_speaker_2.npz b/bark/assets/prompts/pl_speaker_2.npz
diff --git a/bark/assets/prompts/pl_speaker_3.npz b/bark/assets/prompts/pl_speaker_3.npz
diff --git a/bark/assets/prompts/pl_speaker_4.npz b/bark/assets/prompts/pl_speaker_4.npz
diff --git a/bark/assets/prompts/pl_speaker_5.npz b/bark/assets/prompts/pl_speaker_5.npz
diff --git a/bark/assets/prompts/pl_speaker_6.npz b/bark/assets/prompts/pl_speaker_6.npz
diff --git a/bark/assets/prompts/pl_speaker_7.npz b/bark/assets/prompts/pl_speaker_7.npz
diff --git a/bark/assets/prompts/pl_speaker_8.npz b/bark/assets/prompts/pl_speaker_8.npz
diff --git a/bark/assets/prompts/pl_speaker_9.npz b/bark/assets/prompts/pl_speaker_9.npz
diff --git a/bark/assets/prompts/pt_speaker_0.npz b/bark/assets/prompts/pt_speaker_0.npz
diff --git a/bark/assets/prompts/pt_speaker_1.npz b/bark/assets/prompts/pt_speaker_1.npz
diff --git a/bark/assets/prompts/pt_speaker_2.npz b/bark/assets/prompts/pt_speaker_2.npz
diff --git a/bark/assets/prompts/pt_speaker_3.npz b/bark/assets/prompts/pt_speaker_3.npz
diff --git a/bark/assets/prompts/pt_speaker_4.npz b/bark/assets/prompts/pt_speaker_4.npz
diff --git a/bark/assets/prompts/pt_speaker_5.npz b/bark/assets/prompts/pt_speaker_5.npz
diff --git a/bark/assets/prompts/pt_speaker_6.npz b/bark/assets/prompts/pt_speaker_6.npz
diff --git a/bark/assets/prompts/pt_speaker_7.npz b/bark/assets/prompts/pt_speaker_7.npz
diff --git a/bark/assets/prompts/pt_speaker_8.npz b/bark/assets/prompts/pt_speaker_8.npz
diff --git a/bark/assets/prompts/pt_speaker_9.npz b/bark/assets/prompts/pt_speaker_9.npz
diff --git a/bark/assets/prompts/ru_speaker_0.npz b/bark/assets/prompts/ru_speaker_0.npz
diff --git a/bark/assets/prompts/ru_speaker_1.npz b/bark/assets/prompts/ru_speaker_1.npz
diff --git a/bark/assets/prompts/ru_speaker_2.npz b/bark/assets/prompts/ru_speaker_2.npz
diff --git a/bark/assets/prompts/ru_speaker_3.npz b/bark/assets/prompts/ru_speaker_3.npz
diff --git a/bark/assets/prompts/ru_speaker_4.npz b/bark/assets/prompts/ru_speaker_4.npz
diff --git a/bark/assets/prompts/ru_speaker_5.npz b/bark/assets/prompts/ru_speaker_5.npz
diff --git a/bark/assets/prompts/ru_speaker_6.npz b/bark/assets/prompts/ru_speaker_6.npz
diff --git a/bark/assets/prompts/ru_speaker_7.npz b/bark/assets/prompts/ru_speaker_7.npz
diff --git a/bark/assets/prompts/ru_speaker_8.npz b/bark/assets/prompts/ru_speaker_8.npz
diff --git a/bark/assets/prompts/ru_speaker_9.npz b/bark/assets/prompts/ru_speaker_9.npz
diff --git a/bark/assets/prompts/speaker_0.npz b/bark/assets/prompts/speaker_0.npz
diff --git a/bark/assets/prompts/speaker_1.npz b/bark/assets/prompts/speaker_1.npz
diff --git a/bark/assets/prompts/speaker_2.npz b/bark/assets/prompts/speaker_2.npz
diff --git a/bark/assets/prompts/speaker_3.npz b/bark/assets/prompts/speaker_3.npz
diff --git a/bark/assets/prompts/speaker_4.npz b/bark/assets/prompts/speaker_4.npz
diff --git a/bark/assets/prompts/speaker_5.npz b/bark/assets/prompts/speaker_5.npz
diff --git a/bark/assets/prompts/speaker_6.npz b/bark/assets/prompts/speaker_6.npz
diff --git a/bark/assets/prompts/speaker_7.npz b/bark/assets/prompts/speaker_7.npz
diff --git a/bark/assets/prompts/speaker_8.npz b/bark/assets/prompts/speaker_8.npz
diff --git a/bark/assets/prompts/speaker_9.npz b/bark/assets/prompts/speaker_9.npz
diff --git a/bark/assets/prompts/speech_1.npz b/bark/assets/prompts/speech_1.npz
diff --git a/bark/assets/prompts/speech_2.npz b/bark/assets/prompts/speech_2.npz
diff --git a/bark/assets/prompts/speech_3.npz b/bark/assets/prompts/speech_3.npz
diff --git a/bark/assets/prompts/speech_4.npz b/bark/assets/prompts/speech_4.npz
diff --git a/bark/assets/prompts/speech_5.npz b/bark/assets/prompts/speech_5.npz
diff --git a/bark/assets/prompts/speech_6.npz b/bark/assets/prompts/speech_6.npz
diff --git a/bark/assets/prompts/speech_7.npz b/bark/assets/prompts/speech_7.npz
diff --git a/bark/assets/prompts/tr_speaker_0.npz b/bark/assets/prompts/tr_speaker_0.npz
diff --git a/bark/assets/prompts/tr_speaker_1.npz b/bark/assets/prompts/tr_speaker_1.npz
diff --git a/bark/assets/prompts/tr_speaker_2.npz b/bark/assets/prompts/tr_speaker_2.npz
diff --git a/bark/assets/prompts/tr_speaker_3.npz b/bark/assets/prompts/tr_speaker_3.npz
diff --git a/bark/assets/prompts/tr_speaker_4.npz b/bark/assets/prompts/tr_speaker_4.npz
diff --git a/bark/assets/prompts/tr_speaker_5.npz b/bark/assets/prompts/tr_speaker_5.npz
diff --git a/bark/assets/prompts/tr_speaker_6.npz b/bark/assets/prompts/tr_speaker_6.npz
diff --git a/bark/assets/prompts/tr_speaker_7.npz b/bark/assets/prompts/tr_speaker_7.npz
diff --git a/bark/assets/prompts/tr_speaker_8.npz b/bark/assets/prompts/tr_speaker_8.npz
diff --git a/bark/assets/prompts/tr_speaker_9.npz b/bark/assets/prompts/tr_speaker_9.npz
diff --git a/bark/assets/prompts/zh_speaker_0.npz b/bark/assets/prompts/zh_speaker_0.npz
diff --git a/bark/assets/prompts/zh_speaker_1.npz b/bark/assets/prompts/zh_speaker_1.npz
diff --git a/bark/assets/prompts/zh_speaker_2.npz b/bark/assets/prompts/zh_speaker_2.npz
diff --git a/bark/assets/prompts/zh_speaker_3.npz b/bark/assets/prompts/zh_speaker_3.npz
diff --git a/bark/assets/prompts/zh_speaker_4.npz b/bark/assets/prompts/zh_speaker_4.npz
diff --git a/bark/assets/prompts/zh_speaker_5.npz b/bark/assets/prompts/zh_speaker_5.npz
diff --git a/bark/assets/prompts/zh_speaker_6.npz b/bark/assets/prompts/zh_speaker_6.npz
diff --git a/bark/assets/prompts/zh_speaker_7.npz b/bark/assets/prompts/zh_speaker_7.npz
diff --git a/bark/assets/prompts/zh_speaker_8.npz b/bark/assets/prompts/zh_speaker_8.npz
diff --git a/bark/assets/prompts/zh_speaker_9.npz b/bark/assets/prompts/zh_speaker_9.npz
diff --git a/bark/generation.py b/bark/generation.py
@@ -48,10 +48,26 @@ def autocast():
 SAMPLE_RATE = 24_000
 
 
-ALLOWED_PROMPTS = set(
-    [f"speech_{n}" for n in range(8)] +
-    [f"music_{n}" for n in range(6)]
-)
+SUPPORTED_LANGS = [
+    ("English", "en"),
+    ("German", "de"),
+    ("Spanish", "es"),
+    ("French", "fr"),
+    ("Hindi", "hi"),
+    ("Italian", "it"),
+    ("Japanese", "ja"),
+    ("Korean", "ko"),
+    ("Polish", "pl"),
+    ("Portuguese", "pt"),
+    ("Russian", "ru"),
+    ("Turkish", "tr"),
+    ("Chinese", "zh"),
+]
+
+ALLOWED_PROMPTS = {"announcer"}
+for _, lang in SUPPORTED_LANGS:
+    for n in range(10):
+        ALLOWED_PROMPTS.add(f"{lang}_speaker_{n}")
 
 
 logger = logging.getLogger(__name__)
@@ -83,6 +99,13 @@ def autocast():
 }
 
 
+if not hasattr(torch.nn.functional, 'scaled_dot_product_attention'):
+    logger.warning(
+        "torch version does not support flash attention. You will get significantly faster" +
+        " inference speed by upgrade torch to newest version / nightly."
+    )
+
+
 def _string_md5(s):
     m = hashlib.md5()
     m.update(s.encode("utf-8"))

diff --git a/bark/notebooks/fake_classifier.ipynb → notebooks/fake_classifier.ipynb b/bark/notebooks/fake_classifier.ipynb → notebooks/fake_classifier.ipynb