app.py

from transformers import pipeline
import gradio as gr
from transformers import WhisperForConditionalGeneration

pipe = pipeline(model="matteocirca/whisper-small-it")
# pipe = pipeline(model="openai/whisper-small")
# model = WhisperForConditionalGeneration.from_pretrained("matteocirca/whisper-small-it")

segments = []

def audio2segments(audio,word):
    global segments
    if not segments:
        print(pipe(audio).keys())
        # segments = pipe(audio)["segments"]
        # segments = model.transcribe(audio)["segments"]
    elif not word:
        return "No word detected"
    else:
        ranges = []
        for s in segments:
            if word in s['text'].replace(',',' , ').split(" "):
                ranges.append((s['start'],s['end']))
        res = ""
        for i,r in enumerate(ranges):
            res += f"{i}) {r[0]}-{r[1]}\n "
        return res


def find_segment():
    global segments
    res = ""


iface = gr.Interface(
    fn=audio2segments, 
    inputs=[gr.Audio(sources=["microphone"], type="filepath"),"text"],
    outputs="text",
    title="Whisper Small Italian",
    description="Realtime demo for Italian speech recognition using a fine-tuned Whisper small model.",
)


iface.launch()