app.py

import torch
import os
import random
import gradio as gr
from TTS.api import TTS
from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan, pipeline
import base64
from datasets import load_dataset
from diffusers import DiffusionPipeline
from huggingface_hub import login
import numpy as np
import spaces

@spaces.GPU
def guessanImage(model, image):
    imgclassifier  = pipeline("image-classification", model=model)
    if image is not None:  
        description = imgclassifier(image)
    return description

@spaces.GPU
def guessanAge(model, image):
    imgclassifier  = pipeline("image-classification", model=model)
    if image is not None:  
        description = imgclassifier(image)
    return description    

@spaces.GPU(duration=120)
def text2speech(text, no, sample):
    os.environ["COQUI_TOS_AGREED"] = "1"
    if len(text) > 0:
        tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2")
        wav = tts.tts_to_file(text=text, file_path="output.wav", speaker_wav="sampleaudio/abraham.wav", language="en")
        return wav

@spaces.GPU
def ImageGenFromText(text, model):
    api_key = os.getenv("fluxauth")
    login(token=api_key)
    
    if len(text) > 0:
        dtype = torch.bfloat16
        device = "cuda" if torch.cuda.is_available() else "cpu"
        MAX_SEED = np.iinfo(np.int32).max
        seed = random.randint(0, MAX_SEED)
        pipe = DiffusionPipeline.from_pretrained(model, torch_dtype=dtype).to(device)
        generator = torch.Generator().manual_seed(seed)
        image = pipe(
                prompt = text, 
                width = 512,
                height = 512,
                num_inference_steps = 4, 
                generator = generator,
                guidance_scale=0.0
        ).images[0]
        print(image)
        return image


radio1 = gr.Radio(["microsoft/resnet-50", "google/vit-base-patch16-224", "apple/mobilevit-small"], value="microsoft/resnet-50", label="Select a Classifier", info="Image Classifier")
tab1 = gr.Interface(
    fn=guessanImage,
    inputs=[radio1, gr.Image(type="pil")],
    outputs=["text"],
)

radio2 = gr.Radio(["nateraw/vit-age-classifier"], value="nateraw/vit-age-classifier", label="Select an Age Classifier", info="Age Classifier")
tab2 = gr.Interface(
    fn=guessanAge,
    inputs=[radio2, gr.Image(type="pil")],
    outputs=["text"],
)


textbox = gr.Textbox(value="good morning pineapple! looking very good very nice!")
sampletext = gr.HTML("""
<span>If you do not sample your voice my voice will be used as input<span><
 <audio controls autoplay>
    <source src="sampleaudio/abraham.wav" type="audio/wav">
    Your browser does not support the audio element.
</audio>
""")
micinput = gr.Audio(sources=['microphone'], type="filepath", format="wav", value="sampleaudio/abraham.wav")

tab3 = gr.Interface(
    fn=text2speech,
    inputs=[textbox, sampletext, micinput],
    outputs=["audio"],
)

radio4 = gr.Radio(["black-forest-labs/FLUX.1-schnell"], value="black-forest-labs/FLUX.1-schnell", label="Select", info="text to image")
tab4 = gr.Interface(
    fn=ImageGenFromText,
    inputs=["text", radio4],
    outputs=["image"],
)

demo = gr.TabbedInterface([tab1, tab2, tab3, tab4], ["Describe", "Estimage Age", "Speak", "Generate Image"])
demo.launch()