streamlit_app.py

import streamlit as st
import tempfile
from faster_whisper import WhisperModel
import textwrap
from datetime import timedelta
from xml.sax.saxutils import escape

st.set_page_config(page_title="VO to Subtitle Generator Multi-languages version", layout="wide")
st.title("🌍 VO to Subtitle Generator — Multi-languages Version")

def format_time(seconds):
    td = timedelta(seconds=seconds)
    result = str(td)[:11].replace(".", ",")
    return result if "," in result else result + ",000"

def generate_srt(segments, max_chars, max_lines):
    srt_text = ""
    count = 1
    for seg in segments:
        start = seg.start
        end = seg.end
        text = seg.text.strip()
        lines = textwrap.wrap(text, width=max_chars)
        grouped = [lines[i:i+max_lines] for i in range(0, len(lines), max_lines)]
        chunk_count = len(grouped)
        duration = end - start
        chunk_duration = duration / chunk_count if chunk_count > 0 else duration

        for j, chunk in enumerate(grouped):
            chunk_start = start + j * chunk_duration
            chunk_end = chunk_start + chunk_duration
            timestamp = f"{format_time(chunk_start)} --> {format_time(chunk_end)}"
            content = "\n".join(chunk)
            srt_text += f"{count}\n{timestamp}\n{content}\n\n"
            count += 1
    return srt_text

def generate_fcpxml(segments, version):
    xml = [f'<?xml version="1.0" encoding="UTF-8"?>',
           f'<!DOCTYPE fcpxml>',
           f'<fcpxml version="{version}">',
           '  <resources>',
           '    <format id="r1" name="FFVideoFormat1080p25" frameDuration="100/2500s" width="1920" height="1080" colorSpace="1-1-1 (Rec. 709)"/>',
           '    <effect id="r2" name="Custom" uid=".../Titles.localized/Build In:Out.localized/Custom.localized/Custom.moti"/>',
           '  </resources>',
           '  <library>',
           '    <event name="Subtitles">',
           '      <project name="Subtitles">',
           f'        <sequence format="r1" tcStart="0s" tcFormat="NDF">',
           '          <spine>',
           f'            <gap name="Gap" offset="0s" duration="{round(segments[-1].end, 2)}s">']

    for i, seg in enumerate(segments):
        start = round(seg.start, 2)
        duration = round(seg.end - seg.start, 2)
        text = escape(seg.text.strip())
        xml.append(
            f'              <title name="Subtitle {i+1}" lane="1" offset="{start}s" ref="r2" duration="{duration}s">',
            f'                <param name="Position" key="9999/10199/10201/1/100/101" value="0 -415"/>',
            f'                <param name="Alignment" key="9999/10199/10201/2/354/1002961760/401" value="1 (Center)"/>',
            f'                <text>',
            f'                  <text-style ref="ts{i+1}">{text}</text-style>',
            f'                </text>',
            f'                <text-style-def id="ts{i+1}">',
            f'                  <text-style font="Arial" fontSize="60" fontColor="1 1 1 1" alignment="center" bold="1"',
            f'                             strokeColor="0.3 0.3 0.3 1" strokeWidth="-1"',
            f'                             shadowColor="0 0 0 0.75" shadowOffset="3 315" kerning="1.2"/>',
            f'                </text-style-def>',
            f'              </title>'
        )

    xml.extend(['            </gap>',
                '          </spine>',
                '        </sequence>',
                '      </project>',
                '    </event>',
                '  </library>',
                '</fcpxml>'])
    return '\n'.join(xml)

# --- UI Sidebar ---
with st.sidebar:
    st.header("⚙️ Settings")
    uploaded_file = st.file_uploader("Upload MP3 or WAV", type=["mp3", "wav"])
    model_size = st.selectbox("Model Size", ["tiny", "base", "small", "medium"])
    layout = st.selectbox("Video Layout", ["Horizontal (37 chars)", "Vertical (25 chars)"])
    lines = st.selectbox("Lines per Subtitle", [1, 2], index=1)
    language_map = {
        "Auto": None,
        "Arabic": "ar",
        "English": "en",
        "French": "fr",
        "Farsi": "fa",
        "Spanish": "es"
    }
    language = st.selectbox("Language", list(language_map.keys()))
    export_format = st.selectbox("Export Format", ["srt", "fcpxml"])
    fcpxml_version = st.selectbox("FCPXML Version", ["1.13", "1.12", "1.11", "1.8"], index=3) if export_format == "fcpxml" else None

    if 'subtitle_data' not in st.session_state:
        st.session_state.subtitle_data = ""
        st.session_state.text_dir = "rtl"
        st.session_state.generated = False

    if uploaded_file and st.button("🔁 Generate Subtitle"):
        with st.spinner("Transcribing with Whisper..."):
            with tempfile.NamedTemporaryFile(delete=False) as temp_audio:
                temp_audio.write(uploaded_file.read())
                temp_audio.flush()
                whisper = WhisperModel(model_size, device="cpu", compute_type="int8")
                segments_gen, _ = whisper.transcribe(temp_audio.name, language=language_map[language])
                segments = list(segments_gen)
                st.session_state.segments = segments

            max_chars = 25 if "Vertical" in layout else 37
            max_lines = int(lines)

            if export_format == "srt":
                st.session_state.subtitle_data = generate_srt(segments, max_chars, max_lines)
            else:
                st.session_state.subtitle_data = generate_fcpxml(segments, fcpxml_version)

            st.session_state.generated = True

# --- Main UI Output ---
if st.session_state.generated and st.session_state.subtitle_data:
    col1, col2 = st.columns([1, 6])
    with col1:
        st.write("Text Direction:")
        if st.button("⬅️ RTL"):
            st.session_state.text_dir = "rtl"
        if st.button("➡️ LTR"):
            st.session_state.text_dir = "ltr"

    with col2:
        st.markdown("### ✏️ Edit Before Download")
        preview_html = f'''
            <textarea id="subtitle_editor" name="subtitle_editor"
                style="width: 100%; height: 300px; padding: 1.5em 2em;
                border: 1px solid #ccc; border-radius: 8px;
                font-family: monospace; font-size: 14px;
                direction: {st.session_state.text_dir}; white-space: pre-wrap;">{st.session_state.subtitle_data}</textarea>
            <script>
                const editor = document.getElementById('subtitle_editor');
                editor.addEventListener('input', () => {{
                    window.subtitleEdited = editor.value;
                }});
                window.subtitleEdited = editor.value;
            </script>
        '''
        st.components.v1.html(preview_html, height=360)

        st.download_button(
            label="⬇️ Download Subtitle",
            data=st.session_state.subtitle_data,
            file_name="subtitles." + ("srt" if export_format == "srt" else "fcpxml"),
            mime="text/plain"
        )