import streamlit as st import whisper from TTS.api import TTS from moviepy.editor import VideoFileClip, AudioFileClip import os from tempfile import NamedTemporaryFile # Page config st.set_page_config(page_title="AI Voiceover Generator", layout="centered") st.title("๐ŸŽค AI Voiceover + Subtitle Enhancer") # Load models @st.cache_resource def load_whisper_model(): return whisper.load_model("small") @st.cache_resource def load_tts_model(): return TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False, gpu=False) whisper_model = load_whisper_model() tts = load_tts_model() # Upload video video_file = st.file_uploader("Upload a short video clip (MP4 preferred)", type=["mp4", "mov", "avi"]) if video_file: with NamedTemporaryFile(delete=False, suffix=".mp4") as tmp_video: tmp_video.write(video_file.read()) tmp_video_path = tmp_video.name st.video(tmp_video_path) # Extract audio video = VideoFileClip(tmp_video_path) audio_path = tmp_video_path.replace(".mp4", ".wav") video.audio.write_audiofile(audio_path) # Transcribe st.info("Transcribing using Whisper...") result = whisper_model.transcribe(audio_path) st.subheader("๐Ÿ“ Detected Speech") st.write(result["text"]) # User input for voiceover custom_text = st.text_area("Enter your custom voiceover text:", "Hereโ€™s my voiceover explaining the video...") if st.button("Generate AI Voiceover"): voice_output_path = audio_path.replace(".wav", "_ai_voice.wav") tts.tts_to_file(text=custom_text, file_path=voice_output_path) st.audio(voice_output_path) # Replace original audio with new one final_video = video.set_audio(AudioFileClip(voice_output_path)) final_path = tmp_video_path.replace(".mp4", "_final.mp4") final_video.write_videofile(final_path, codec="libx264", audio_codec="aac") with open(final_path, "rb") as f: st.download_button(label="๐Ÿ“ฅ Download Final Video", data=f, file_name="final_ai_video.mp4")