app.py

from transformers import pipeline
import gradio as gr
from numpy import random 
from PIL import Image
pipe = pipeline(model="fimster/whisper-small-sv-SE")  # change to "your-username/the-name-you-picked"
images = ["katt", "melon", "banan"]
image = random.choice(images)

def transcribe(audio, img):
    text = pipe(audio)["text"]
    text = text.replace("!", "")
    text = text.replace(".", "")
    text = text.replace(",", "")
    returntext = ""
    if text.lower() != image.lower():
        returntext = "Du svarade fel, ditt svar var: " + text + ", rätt svar var: " + image
    else:
        returntext = "Du hade rätt, svaret var: " + image
    return returntext

iface = gr.Interface(
    fn=transcribe, 
    inputs=[gr.Audio(source="microphone", type="filepath"), gr.Image("./images/" + image + ".jpeg")], 
    outputs="text",
    title="Whisper Small Swedish",
    description="Proof of concept for Duolingo-type gamified language learning. Record and say what the thing in the picture is, then click submit and the model will check if you were correct!",
)

iface.launch()