Update app.py
Browse files
app.py
CHANGED
@@ -22,7 +22,6 @@ generate_kwargs = {
|
|
22 |
"max_new_tokens": 64,
|
23 |
}
|
24 |
|
25 |
-
|
26 |
model_dict = {
|
27 |
"whisper-large-v3-turbo": "openai/whisper-large-v3-turbo",
|
28 |
"kotoba-whisper-v2.0": "kotoba-tech/kotoba-whisper-v2.0",
|
@@ -71,13 +70,12 @@ def transcribe_common(audio: str, model: str) -> str:
|
|
71 |
logger.success(f"Finished in {end_time - start_time:.2f}s\n{result}")
|
72 |
|
73 |
# Guardar resultado en un archivo .str
|
74 |
-
output_path = f"{Path(filename).stem}.str"
|
75 |
with open(output_path, "w", encoding="utf-8") as f:
|
76 |
f.write(result)
|
77 |
|
78 |
logger.info(f"Transcription saved to {output_path}")
|
79 |
-
return
|
80 |
-
|
81 |
|
82 |
|
83 |
def transcribe_others(audio) -> tuple[str, str]:
|
@@ -92,7 +90,7 @@ def transcribe_anime_whisper(audio) -> str:
|
|
92 |
|
93 |
initial_md = """
|
94 |
# Anime-Whisper Demo
|
95 |
-
[**Anime Whisper**](https://huggingface.co/litagin/anime-whisper): 5千時間以上のアニメ調セリフと台本でファインチューニングされた日本語音声認識モデルのデモです。句読点や感嘆符がリズムや感情に合わせて自然に付き、NSFW
|
96 |
- デモでは**音声は15秒まで**しか受け付けません
|
97 |
- 日本語のみ対応 (Japanese only)
|
98 |
- 比較のために [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) と [kotoba-tech/kotoba-whisper-v2.0](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0) も用意しています
|
@@ -115,16 +113,16 @@ with gr.Blocks() as app:
|
|
115 |
with gr.Column():
|
116 |
gr.Markdown("### Anime-Whisper")
|
117 |
button_galgame = gr.Button("Transcribe with Anime-Whisper")
|
118 |
-
output_galgame = gr.
|
119 |
gr.Markdown("### Comparison")
|
120 |
button_others = gr.Button("Transcribe with other models")
|
121 |
with gr.Row():
|
122 |
with gr.Column():
|
123 |
gr.Markdown("### Whisper-Large-V3-Turbo")
|
124 |
-
output_v3 = gr.
|
125 |
with gr.Column():
|
126 |
gr.Markdown("### Kotoba-Whisper-V2.0")
|
127 |
-
output_kotoba_v2 = gr.
|
128 |
|
129 |
button_galgame.click(
|
130 |
transcribe_anime_whisper,
|
@@ -138,3 +136,4 @@ with gr.Blocks() as app:
|
|
138 |
)
|
139 |
|
140 |
app.launch(inbrowser=True)
|
|
|
|
22 |
"max_new_tokens": 64,
|
23 |
}
|
24 |
|
|
|
25 |
model_dict = {
|
26 |
"whisper-large-v3-turbo": "openai/whisper-large-v3-turbo",
|
27 |
"kotoba-whisper-v2.0": "kotoba-tech/kotoba-whisper-v2.0",
|
|
|
70 |
logger.success(f"Finished in {end_time - start_time:.2f}s\n{result}")
|
71 |
|
72 |
# Guardar resultado en un archivo .str
|
73 |
+
output_path = f"{Path(filename).stem}_{model}.str"
|
74 |
with open(output_path, "w", encoding="utf-8") as f:
|
75 |
f.write(result)
|
76 |
|
77 |
logger.info(f"Transcription saved to {output_path}")
|
78 |
+
return output_path # Devuelve el path del archivo transcrito
|
|
|
79 |
|
80 |
|
81 |
def transcribe_others(audio) -> tuple[str, str]:
|
|
|
90 |
|
91 |
initial_md = """
|
92 |
# Anime-Whisper Demo
|
93 |
+
[**Anime Whisper**](https://huggingface.co/litagin/anime-whisper): 5千時間以上のアニメ調セリフと台本でファインチューニングされた日本語音声認識モデルのデモです。句読点や感嘆符がリズムや感情に合わせて自然に付き、NSFW含む非言語発言もうまく台本調に書き起こされます。
|
94 |
- デモでは**音声は15秒まで**しか受け付けません
|
95 |
- 日本語のみ対応 (Japanese only)
|
96 |
- 比較のために [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) と [kotoba-tech/kotoba-whisper-v2.0](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0) も用意しています
|
|
|
113 |
with gr.Column():
|
114 |
gr.Markdown("### Anime-Whisper")
|
115 |
button_galgame = gr.Button("Transcribe with Anime-Whisper")
|
116 |
+
output_galgame = gr.File(label="Download Anime-Whisper Transcription")
|
117 |
gr.Markdown("### Comparison")
|
118 |
button_others = gr.Button("Transcribe with other models")
|
119 |
with gr.Row():
|
120 |
with gr.Column():
|
121 |
gr.Markdown("### Whisper-Large-V3-Turbo")
|
122 |
+
output_v3 = gr.File(label="Download Whisper-Large-V3-Turbo Transcription")
|
123 |
with gr.Column():
|
124 |
gr.Markdown("### Kotoba-Whisper-V2.0")
|
125 |
+
output_kotoba_v2 = gr.File(label="Download Kotoba-Whisper-V2.0 Transcription")
|
126 |
|
127 |
button_galgame.click(
|
128 |
transcribe_anime_whisper,
|
|
|
136 |
)
|
137 |
|
138 |
app.launch(inbrowser=True)
|
139 |
+
|