Update app.py

app.py

@@ -26,6 +26,7 @@ speech2lang = Speech2Lang.from_pretrained(
     quantize_dtype="float16",
 )
 
+
 iso_codes = ['abk', 'afr', 'amh', 'ara', 'asm', 'ast', 'aze', 'bak', 'bas', 'bel', 'ben', 'bos', 'bre', 'bul', 'cat', 'ceb', 'ces', 'chv', 'ckb', 'cmn', 'cnh', 'cym', 'dan', 'deu', 'dgd', 'div', 'ell', 'eng', 'epo', 'est', 'eus', 'fas', 'fil', 'fin', 'fra', 'frr', 'ful', 'gle', 'glg', 'grn', 'guj', 'hat', 'hau', 'heb', 'hin', 'hrv', 'hsb', 'hun', 'hye', 'ibo', 'ina', 'ind', 'isl', 'ita', 'jav', 'jpn', 'kab', 'kam', 'kan', 'kat', 'kaz', 'kea', 'khm', 'kin', 'kir', 'kmr', 'kor', 'lao', 'lav', 'lga', 'lin', 'lit', 'ltz', 'lug', 'luo', 'mal', 'mar', 'mas', 'mdf', 'mhr', 'mkd', 'mlt', 'mon', 'mri', 'mrj', 'mya', 'myv', 'nan', 'nep', 'nld', 'nno', 'nob', 'npi', 'nso', 'nya', 'oci', 'ori', 'orm', 'ory', 'pan', 'pol', 'por', 'pus', 'quy', 'roh', 'ron', 'rus', 'sah', 'sat', 'sin', 'skr', 'slk', 'slv', 'sna', 'snd', 'som', 'sot', 'spa', 'srd', 'srp', 'sun', 'swa', 'swe', 'swh', 'tam', 'tat', 'tel', 'tgk', 'tgl', 'tha', 'tig', 'tir', 'tok', 'tpi', 'tsn', 'tuk', 'tur', 'twi', 'uig', 'ukr', 'umb', 'urd', 'uzb', 'vie', 'vot', 'wol', 'xho', 'yor', 'yue', 'zho', 'zul']
 lang_names = ['Abkhazian', 'Afrikaans', 'Amharic', 'Arabic', 'Assamese', 'Asturian', 'Azerbaijani', 'Bashkir', 'Basa (Cameroon)', 'Belarusian', 'Bengali', 'Bosnian', 'Breton', 'Bulgarian', 'Catalan', 'Cebuano', 'Czech', 'Chuvash', 'Central Kurdish', 'Mandarin Chinese', 'Hakha Chin', 'Welsh', 'Danish', 'German', 'Dagaari Dioula', 'Dhivehi', 'Modern Greek (1453-)', 'English', 'Esperanto', 'Estonian', 'Basque', 'Persian', 'Filipino', 'Finnish', 'French', 'Northern Frisian', 'Fulah', 'Irish', 'Galician', 'Guarani', 'Gujarati', 'Haitian', 'Hausa', 'Hebrew', 'Hindi', 'Croatian', 'Upper Sorbian', 'Hungarian', 'Armenian', 'Igbo', 'Interlingua (International Auxiliary Language Association)', 'Indonesian', 'Icelandic', 'Italian', 'Javanese', 'Japanese', 'Kabyle', 'Kamba (Kenya)', 'Kannada', 'Georgian', 'Kazakh', 'Kabuverdianu', 'Khmer', 'Kinyarwanda', 'Kirghiz', 'Northern Kurdish', 'Korean', 'Lao', 'Latvian', 'Lungga', 'Lingala', 'Lithuanian', 'Luxembourgish', 'Ganda', 'Luo (Kenya and Tanzania)', 'Malayalam', 'Marathi', 'Masai', 'Moksha', 'Eastern Mari', 'Macedonian', 'Maltese', 'Mongolian', 'Maori', 'Western Mari', 'Burmese', 'Erzya', 'Min Nan Chinese', 'Nepali (macrolanguage)', 'Dutch', 'Norwegian Nynorsk', 'Norwegian Bokmål', 'Nepali (individual language)', 'Pedi', 'Nyanja', 'Occitan (post 1500)', 'Oriya (macrolanguage)', 'Oromo', 'Odia', 'Panjabi', 'Polish', 'Portuguese', 'Pushto', 'Ayacucho Quechua', 'Romansh', 'Romanian', 'Russian', 'Yakut', 'Santali', 'Sinhala', 'Saraiki', 'Slovak', 'Slovenian', 'Shona', 'Sindhi', 'Somali', 'Southern Sotho', 'Spanish', 'Sardinian', 'Serbian', 'Sundanese', 'Swahili (macrolanguage)', 'Swedish', 'Swahili (individual language)', 'Tamil', 'Tatar', 'Telugu', 'Tajik', 'Tagalog', 'Thai', 'Tigre', 'Tigrinya', 'Toki Pona', 'Tok Pisin', 'Tswana', 'Turkmen', 'Turkish', 'Twi', 'Uighur', 'Ukrainian', 'Umbundu', 'Urdu', 'Uzbek', 'Vietnamese', 'Votic', 'Wolof', 'Xhosa', 'Yoruba', 'Yue Chinese', 'Chinese', 'Zulu']
 
@@ -109,6 +110,35 @@ def predict(audio_path, src_lang: str, task: str, beam_size, long_form: bool, te
     return code2lang[lang_code], text
 
 
+_DESCRIPTION=r'''
+OWSM is an Open Whisper-style Speech Model from [CMU WAVLab](https://www.wavlab.org/).
+It reproduces Whisper-style training using publicly available data and an open-source toolkit [ESPnet](https://github.com/espnet/espnet).
+
+OWSM v3 is trained on 180k hours of paired speech data. It supports various speech-to-text tasks:
+- Speech recognition for 151 languages
+- Any-to-any language speech translation
+- Timestamp prediction
+- Long-form transcription
+- Language identification
+
+For more details about OWSM, please check out our paper [here](https://arxiv.org/abs/2309.13876) (ASRU 2023).
+
+```
+@misc{peng2023reproducing,
+      title={Reproducing Whisper-Style Training Using an Open-Source Toolkit and Publicly Available Data}, 
+      author={Yifan Peng and Jinchuan Tian and Brian Yan and Dan Berrebbi and Xuankai Chang and Xinjian Li and Jiatong Shi and Siddhant Arora and William Chen and Roshan Sharma and Wangyou Zhang and Yui Sudo and Muhammad Shakeel and Jee-weon Jung and Soumi Maiti and Shinji Watanabe},
+      year={2023},
+      eprint={2309.13876},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL}
+}
+```
+
+Disclaimer: OWSM has not been thoroughly evaluated in all tasks. Due to limited training data, it may not perform well for certain language directions.
+
+
+'''
+
 demo = gr.Interface(
     predict,
     inputs=[
@@ -123,6 +153,8 @@ demo = gr.Interface(
         gr.Text(label="Predicted Language", info="Language identification is performed if language is unknown."),
         gr.Text(label="Predicted Text", info="Best hypothesis, without timestamps."),
     ],
+    title="Demo of OWSM v3: An Open Whisper-style Speech Model from CMU WAVLab",
+    description=_DESCRIPTION,
 )