Hugging Face's logo

Spaces:
justREE
/
StoryTeller_Assignment
Running

App Files Community
StoryTeller_Assignment
File size: 4,885 Bytes 
434afff
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
 
# app.py

import io
import wave
import re
import streamlit as st
from transformers import pipeline, SpeechT5Processor, SpeechT5HifiGan
from datasets import load_dataset
from PIL import Image
import numpy as np
import torch

# ─────────────────────────────────────────────────────────────
# 1) LOAD PIPELINES
# ─────────────────────────────────────────────────────────────
@st.cache_resource(show_spinner=False)
def load_captioner():
    return pipeline("image-to-text", model="Salesforce/blip-image-captioning-base", device="cpu")

@st.cache_resource(show_spinner=False)
def load_story_generator():
    return pipeline("text-generation", model="microsoft/Phi-4-mini-reasoning", device="cpu")

@st.cache_resource(show_spinner=False)
def load_tts_pipe():
    processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
    model = pipeline("text-to-speech", model="microsoft/speecht5_tts", device="cpu")
    vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
    speaker_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
    speaker_embedding = torch.tensor(speaker_dataset[7306]["xvector"]).unsqueeze(0)
    return processor, model, vocoder, speaker_embedding

# ─────────────────────────────────────────────────────────────
# 2) PIPELINE FUNCTIONS
# ─────────────────────────────────────────────────────────────
def get_caption(image, captioner):
    return captioner(image)[0]['generated_text']

def generate_story(caption, generator):
    prompt = f"Write a short, magical story for children aged 3 to 10 based on this scene: {caption}. Keep it under 100 words."
    outputs = generator(
        prompt,
        max_new_tokens=120,
        temperature=0.8,
        top_p=0.95,
        do_sample=True
    )
    story = outputs[0]["generated_text"]
    return clean_story_output(story, prompt)

def clean_story_output(story, prompt):
    story = story[len(prompt):].strip() if story.startswith(prompt) else story
    if "." in story:
        story = story[: story.rfind(".") + 1]
    return sentence_case(story)

def sentence_case(text):
    parts = re.split(r'([.!?])', text)
    out = []
    for i in range(0, len(parts) - 1, 2):
        sentence = parts[i].strip().capitalize()
        out.append(f"{sentence}{parts[i + 1]}")
    if len(parts) % 2:
        last = parts[-1].strip().capitalize()
        if last:
            out.append(last)
    return " ".join(out)

def convert_to_audio(text, processor, tts_pipe, vocoder, speaker_embedding):
    inputs = processor(text=text, return_tensors="pt")
    speech = tts_pipe.model.generate_speech(inputs["input_ids"], speaker_embedding, vocoder=vocoder)
    pcm = (speech.numpy() * 32767).astype(np.int16)
    buffer = io.BytesIO()
    with wave.open(buffer, "wb") as wf:
        wf.setnchannels(1)
        wf.setsampwidth(2)
        wf.setframerate(16000)
        wf.writeframes(pcm.tobytes())
    buffer.seek(0)
    return buffer.read()

# ─────────────────────────────────────────────────────────────
# 3) STREAMLIT APP UI
# ─────────────────────────────────────────────────────────────
st.set_page_config(page_title="Magic Storyteller", layout="centered")
st.title("🧚 Magic Storyteller")
st.markdown("Upload an image to generate a magical story and hear it read aloud!")

uploaded = st.file_uploader("Upload Image", type=["jpg", "jpeg", "png"])
if uploaded:
    image = Image.open(uploaded)
    st.image(image, caption="Your uploaded image", use_column_width=True)

    st.subheader("🖼️ Step 1: Captioning")
    captioner = load_captioner()
    caption = get_caption(image, captioner)
    st.markdown(f"**Caption:** {sentence_case(caption)}")

    st.subheader("📖 Step 2: Story Generation")
    story_pipe = load_story_generator()
    story = generate_story(caption, story_pipe)
    st.write(story)

    st.subheader("🔊 Step 3: Listen to the Story")
    processor, tts_pipe, vocoder, speaker_embedding = load_tts_pipe()
    audio_bytes = convert_to_audio(story, processor, tts_pipe, vocoder, speaker_embedding)
    st.audio(audio_bytes, format="audio/wav")
    st.balloons()
else:
    st.info("Please upload an image to begin.")