""" Caption Renderer V4 - Gradio App Converts JSON transcripts to WebM video with styled captions """ import os import json import uuid import tempfile from typing import List, Dict, Optional import gradio as gr import cloudinary import cloudinary.uploader from canvas_renderer import render_frame, WIDTH, HEIGHT from video_encoder import encode_frames_pipe # Configuration CLOUD_NAME = os.environ.get("CLOUDINARY_CLOUD_NAME", "dgfhhszx8") UPLOAD_PRESET = os.environ.get("CLOUDINARY_UPLOAD_PRESET", "testing") FPS = 24 # Frames per second def parse_transcript(transcript_json: str) -> List[Dict]: """Parse the transcript JSON input""" try: data = json.loads(transcript_json) # Handle both direct array and wrapped format if isinstance(data, list): return data elif isinstance(data, dict) and "fullTranscript" in data: return data["fullTranscript"] else: raise ValueError("Invalid transcript format") except json.JSONDecodeError as e: raise ValueError(f"Invalid JSON: {e}") def group_words_into_phrases(transcript: List[Dict], max_words: int = 4, max_duration: float = 2.0) -> List[Dict]: """ Group individual words into display phrases. Each phrase will be shown together, with word-by-word highlighting. Returns list of phrases with structure: { 'words': ['word1', 'word2', ...], 'timings': [{'start': 0.0, 'end': 0.5}, ...], 'start': phrase_start, 'end': phrase_end } """ if not transcript: return [] phrases = [] current_phrase = {'words': [], 'timings': [], 'start': None, 'end': None} for item in transcript: word = item['text'] start = float(item['start']) end = float(item['end']) if current_phrase['start'] is None: current_phrase['start'] = start current_phrase['words'].append(word) current_phrase['timings'].append({'start': start, 'end': end}) current_phrase['end'] = end # Check if we should start a new phrase phrase_duration = current_phrase['end'] - current_phrase['start'] if len(current_phrase['words']) >= max_words or phrase_duration >= max_duration: phrases.append(current_phrase) current_phrase = {'words': [], 'timings': [], 'start': None, 'end': None} # Add remaining words if current_phrase['words']: phrases.append(current_phrase) return phrases def generate_video(transcript_json: str, style: str, progress=gr.Progress()) -> tuple: """ Main video generation function. Args: transcript_json: JSON string with transcript data style: Caption style (hormozi, cinematic, netflix, neon) progress: Gradio progress tracker Returns: Tuple of (video_path, cloudinary_url) """ progress(0, desc="Parsing transcript...") # Parse input try: transcript = parse_transcript(transcript_json) except ValueError as e: raise gr.Error(f"Failed to parse transcript: {e}") if not transcript: raise gr.Error("Empty transcript provided") progress(0.1, desc="Grouping words into phrases...") # Group words into phrases phrases = group_words_into_phrases(transcript) if not phrases: raise gr.Error("No phrases generated from transcript") # Calculate total duration total_duration = max(p['end'] for p in phrases) + 0.5 # Add buffer total_frames = int(total_duration * FPS) progress(0.2, desc=f"Generating {total_frames} frames...") # Generate frames frames = [] for frame_idx in range(total_frames): current_time = frame_idx / FPS # Find which phrase is active at this time active_phrase = None for phrase in phrases: if phrase['start'] <= current_time <= phrase['end']: active_phrase = phrase break if active_phrase: words = active_phrase['words'] # Find which word is active within the phrase active_word_idx = -1 for i, timing in enumerate(active_phrase['timings']): if timing['start'] <= current_time <= timing['end']: active_word_idx = i break frame = render_frame(words, active_word_idx, style) else: # No active phrase - render empty green screen or last phrase dimmed if phrases: last_phrase = phrases[-1] frame = render_frame(last_phrase['words'], -1, style) else: from PIL import Image frame = Image.new('RGB', (WIDTH, HEIGHT), (0, 255, 0)) frames.append(frame) # Update progress every 10% if frame_idx % max(1, total_frames // 10) == 0: pct = 0.2 + (frame_idx / total_frames) * 0.5 progress(pct, desc=f"Rendering frame {frame_idx}/{total_frames}...") progress(0.7, desc="Encoding video...") # Create temp output file output_dir = tempfile.mkdtemp(prefix="caption_video_") output_path = os.path.join(output_dir, f"caption_{uuid.uuid4().hex[:8]}.webm") # Encode to WebM try: encode_frames_pipe(frames, output_path, fps=FPS) except RuntimeError as e: raise gr.Error(f"Video encoding failed: {e}") # Verify output if not os.path.exists(output_path) or os.path.getsize(output_path) < 1000: raise gr.Error("Video encoding produced empty or invalid file") progress(0.85, desc="Uploading to Cloudinary...") # Upload to Cloudinary try: result = cloudinary.uploader.unsigned_upload( output_path, UPLOAD_PRESET, cloud_name=CLOUD_NAME, resource_type="video" ) cloudinary_url = result.get("secure_url", "") except Exception as e: cloudinary_url = f"Upload failed: {e}" progress(1.0, desc="Done!") return output_path, cloudinary_url # Sample transcript for testing SAMPLE_TRANSCRIPT = json.dumps([ {"text": "WATCH", "start": 0.0, "end": 0.5}, {"text": "THIS", "start": 0.5, "end": 1.0}, {"text": "AMAZING", "start": 1.0, "end": 1.6}, {"text": "VIDEO", "start": 1.6, "end": 2.2}, ], indent=2) # Gradio UI with gr.Blocks(title="Caption Renderer V4", theme=gr.themes.Soft()) as demo: gr.Markdown("# 🎬 Caption Renderer V4") gr.Markdown("Convert JSON transcripts to WebM videos with animated captions (green screen)") with gr.Row(): with gr.Column(): transcript_input = gr.Textbox( label="Transcript JSON", placeholder='[{"text": "HELLO", "start": 0.0, "end": 0.5}, ...]', lines=12, value=SAMPLE_TRANSCRIPT ) style_dropdown = gr.Dropdown( choices=["hormozi", "cinematic", "netflix", "neon"], value="hormozi", label="Caption Style" ) generate_btn = gr.Button("🎥 Generate Video", variant="primary") with gr.Column(): video_output = gr.Video(label="Generated Video") cloudinary_url = gr.Textbox(label="Cloudinary URL", interactive=False) generate_btn.click( fn=generate_video, inputs=[transcript_input, style_dropdown], outputs=[video_output, cloudinary_url] ) gr.Markdown("---") gr.Markdown("### Supported Styles") gr.Markdown(""" - **Hormozi**: Gold highlighted word, white inactive, pop animation - **Cinematic**: Premium white/gray with cyan glow - **Netflix**: Netflix red active, white inactive - **Neon**: Magenta/Cyan neon glow effect """) if __name__ == "__main__": demo.launch(server_name="0.0.0.0", server_port=7860, share=False, show_error=True)