app.py

import streamlit as st
import cv2
import numpy as np
import mediapipe as mp
import requests
from gtts import gTTS
import tempfile
import time

# ---------------- CONFIG ----------------
st.set_page_config(page_title="Hand2Voice", layout="wide")

GESTURE_URL = "https://raw.githubusercontent.com/YOUR_USERNAME/Hand2Voice/main/gestures/gesture_rules.json"

# ---------------- SESSION STATE ----------------
if "last_gesture" not in st.session_state:
    st.session_state.last_gesture = ""

if "last_spoken" not in st.session_state:
    st.session_state.last_spoken = ""

# ---------------- LOAD GESTURES ----------------
@st.cache_data
def load_gestures():
    return requests.get(GESTURE_URL).json()["gestures"]

gesture_rules = load_gestures()

# ---------------- MEDIAPIPE ----------------
mp_hands = mp.solutions.hands
hands = mp_hands.Hands(
    static_image_mode=True,
    max_num_hands=1,
    min_detection_confidence=0.7
)

# ---------------- FINGER LOGIC ----------------
def get_finger_states(hand_landmarks):
    finger_tips = [4, 8, 12, 16, 20]
    finger_bases = [2, 6, 10, 14, 18]

    states = []

    # Thumb
    states.append(
        1 if hand_landmarks.landmark[4].x >
             hand_landmarks.landmark[3].x else 0
    )

    # Other fingers
    for tip, base in zip(finger_tips[1:], finger_bases[1:]):
        states.append(
            1 if hand_landmarks.landmark[tip].y <
                 hand_landmarks.landmark[base].y else 0
        )

    return states

# ---------------- MATCH ----------------
def match_gesture(states):
    for name, info in gesture_rules.items():
        if states == info["pattern"]:
            return name
    return "Unknown"

# ---------------- RECOGNIZER ----------------
def recognize(frame):
    rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    result = hands.process(rgb)

    if result.multi_hand_landmarks:
        for hand_landmarks in result.multi_hand_landmarks:
            states = get_finger_states(hand_landmarks)
            return match_gesture(states)

    return "No Hand"

# ---------------- TEXT TO SPEECH ----------------
def speak(text):
    tts = gTTS(text=text, lang="en")
    file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
    tts.save(file.name)
    return file.name

# ---------------- UI ----------------
st.markdown(
    "<h1 style='text-align:center;'>🤟 Hand2Voice</h1>"
    "<p style='text-align:center;'>Real-Time Hand Gesture to Voice</p>",
    unsafe_allow_html=True
)

st.divider()

col1, col2 = st.columns(2)

# -------- CAMERA --------
with col1:
    st.subheader("📷 Live Camera")
    image = st.camera_input("Live gesture feed")

# -------- OUTPUT --------
with col2:
    st.subheader("📝 Live Output")

    if image:
        img_bytes = image.getvalue()
        np_img = np.frombuffer(img_bytes, np.uint8)
        frame = cv2.imdecode(np_img, cv2.IMREAD_COLOR)

        gesture = recognize(frame)
        st.session_state.last_gesture = gesture

        st.markdown(
            f"""
            <div style="
                padding:20px;
                background-color:#dcfce7;
                border-radius:12px;
                font-size:24px;
                font-weight:bold;
                text-align:center;">
                🔊 {gesture}
            </div>
            """,
            unsafe_allow_html=True
        )

        # Speak only if gesture changes
        if gesture not in ["Unknown", "No Hand"] and gesture != st.session_state.last_spoken:
            audio = speak(gesture)
            st.audio(audio)
            st.session_state.last_spoken = gesture

        # Auto refresh (real-time effect)
        time.sleep(0.5)
        st.experimental_rerun()

    else:
        st.info("Enable camera and show hand gestures")