Create new file

app.py

@@ -0,0 +1,196 @@
+import os
+os.system("pip install git+https://github.com/openai/whisper.git")
+
+
+import gradio as gr
+import whisper
+import io
+import os
+import numpy as np
+from datetime import datetime
+
+import assets
+
+def sendToWhisper(audio_record, audio_upload, task, models_selected, language_toggle, language_selected, without_timestamps):
+    results = []
+
+    audio = None
+    if audio_record is not None:
+        audio = audio_record
+    elif audio_upload is not None:
+        audio = audio_upload
+    else:
+        return [["Invalid input"]*5]
+
+    audio = whisper.load_audio(audio)
+    audio = whisper.pad_or_trim(audio)
+
+    for model_name in models_selected:
+        start = datetime.now()
+        model = whisper.load_model(model_name)
+        mel = whisper.log_mel_spectrogram(audio).to(model.device)
+        options = whisper.DecodingOptions(fp16 = False, without_timestamps=without_timestamps, task=task)
+        if language_toggle:
+            options = whisper.DecodingOptions(fp16 = False, without_timestamps=without_timestamps, task=task, language=language_selected)
+        language = ""
+        prob = 0
+        if model_name in assets.lang_detect:
+            _, probs = model.detect_language(mel)
+            language = max(probs, key=probs.get)
+            prob = probs[language]
+        else:
+            language="en"
+            options = whisper.DecodingOptions(fp16 = False, without_timestamps=without_timestamps, task=task, language="en")
+        output_text = whisper.decode(model, mel, options)
+        results.append([model_name, output_text.text, language, str(prob), str((datetime.now() - start).total_seconds())])
+    return results
+
+avail_models = whisper.available_models()
+
+
+
+with gr.Blocks(css=assets.css) as demo:
+    gr.Markdown("This is a demo to use Open AI's Speech to Text (ASR) Model: Whisper. Learn more about the models here on [Github](https://github.com/openai/whisper/search?q=DecodingOptions&type=) FYI: The larger models take a lot longer to transcribe the text :)")
+    gr.Markdown("Here are sample audio files to try out: [Sample Audio](https://drive.google.com/drive/folders/1qYek06ZVeKr9f5Jf35eqi-9CnjNIp98u?usp=sharing)")
+    gr.Markdown("Built by:[@davidtsong](https://twitter.com/davidtsong)")
+    
+    # with gr.Row():
+    with gr.Column():
+
+        # with gr.Column():
+        gr.Markdown("## Input")
+
+        with gr.Row():
+            audio_record = gr.Audio(source="microphone", label="Audio to transcribe", type="filepath",elem_id="audio_inputs")
+            audio_upload = gr.Audio(source="upload", type="filepath", interactive=True,elem_id="audio_inputs")
+
+        models_selected = gr.CheckboxGroup(avail_models, label="Models to use")
+        with gr.Accordion("Settings", open=False):
+            task = gr.Dropdown(["transcribe", "translate"], label="Task", value="transcribe")
+            language_toggle = gr.Dropdown(["Automatic", "Manual"], label="Language Selection", value="Automatic")
+            language_selected = gr.Dropdown(list(assets.LANGUAGES.keys()), label="Language")
+            without_timestamps = gr.Checkbox(label="Without timestamps",value=True)
+        submit = gr.Button(label="Run")
+        
+    # with gr.Row():
+        # with gr.Column():
+        gr.Markdown("## Output")
+        output = gr.Dataframe(headers=["Model", "Text", "Language", "Language Confidence","Time(s)"], label="Results", wrap=True)
+
+    submit.click(fn=sendToWhisper, inputs=[audio_record, audio_upload, task, models_selected, language_toggle, language_selected, without_timestamps], outputs=output)
+
+demo.launch()   
+
+
+assets.py
+
+LANGUAGES = {
+    "en": "english",
+    "zh": "chinese",
+    "de": "german",
+    "es": "spanish",
+    "ru": "russian",
+    "ko": "korean",
+    "fr": "french",
+    "ja": "japanese",
+    "pt": "portuguese",
+    "tr": "turkish",
+    "pl": "polish",
+    "ca": "catalan",
+    "nl": "dutch",
+    "ar": "arabic",
+    "sv": "swedish",
+    "it": "italian",
+    "id": "indonesian",
+    "hi": "hindi",
+    "fi": "finnish",
+    "vi": "vietnamese",
+    "iw": "hebrew",
+    "uk": "ukrainian",
+    "el": "greek",
+    "ms": "malay",
+    "cs": "czech",
+    "ro": "romanian",
+    "da": "danish",
+    "hu": "hungarian",
+    "ta": "tamil",
+    "no": "norwegian",
+    "th": "thai",
+    "ur": "urdu",
+    "hr": "croatian",
+    "bg": "bulgarian",
+    "lt": "lithuanian",
+    "la": "latin",
+    "mi": "maori",
+    "ml": "malayalam",
+    "cy": "welsh",
+    "sk": "slovak",
+    "te": "telugu",
+    "fa": "persian",
+    "lv": "latvian",
+    "bn": "bengali",
+    "sr": "serbian",
+    "az": "azerbaijani",
+    "sl": "slovenian",
+    "kn": "kannada",
+    "et": "estonian",
+    "mk": "macedonian",
+    "br": "breton",
+    "eu": "basque",
+    "is": "icelandic",
+    "hy": "armenian",
+    "ne": "nepali",
+    "mn": "mongolian",
+    "bs": "bosnian",
+    "kk": "kazakh",
+    "sq": "albanian",
+    "sw": "swahili",
+    "gl": "galician",
+    "mr": "marathi",
+    "pa": "punjabi",
+    "si": "sinhala",
+    "km": "khmer",
+    "sn": "shona",
+    "yo": "yoruba",
+    "so": "somali",
+    "af": "afrikaans",
+    "oc": "occitan",
+    "ka": "georgian",
+    "be": "belarusian",
+    "tg": "tajik",
+    "sd": "sindhi",
+    "gu": "gujarati",
+    "am": "amharic",
+    "yi": "yiddish",
+    "lo": "lao",
+    "uz": "uzbek",
+    "fo": "faroese",
+    "ht": "haitian creole",
+    "ps": "pashto",
+    "tk": "turkmen",
+    "nn": "nynorsk",
+    "mt": "maltese",
+    "sa": "sanskrit",
+    "lb": "luxembourgish",
+    "my": "myanmar",
+    "bo": "tibetan",
+    "tl": "tagalog",
+    "mg": "malagasy",
+    "as": "assamese",
+    "tt": "tatar",
+    "haw": "hawaiian",
+    "ln": "lingala",
+    "ha": "hausa",
+    "ba": "bashkir",
+    "jw": "javanese",
+    "su": "sundanese",
+}
+
+lang_detect = ['tiny', 'base', 'small', 'medium', 'large']
+
+css = """
+#audio_inputs{
+    height:100px;
+    max-height:100px;
+}
+"""