Spaces:
Running
Running
| import re | |
| import math | |
| MAX_SCRIPT_LENGTH = 10000 # characters | |
| TTS_1_HD_COST_PER_CHAR = 0.00003 # $30 / 1M chars | |
| GPT_4O_MINI_TTS_COST_PER_SECOND = 0.015 / 60 # $0.015 / minute | |
| CHARS_PER_SECOND_ESTIMATE = 10 # Rough estimate for TTS duration | |
| def parse_dialogue_script(script_text): | |
| """ | |
| Parses a dialogue script into a list of (index, speaker, utterance) tuples. | |
| Input format: "[Speaker] Utterance" per line. | |
| Lines not matching the format are attempted to be parsed as "[Default] Utterance". | |
| """ | |
| lines = script_text.strip().split('\n') | |
| parsed_lines = [] | |
| total_chars = 0 | |
| if len(script_text) > MAX_SCRIPT_LENGTH: | |
| raise ValueError(f"Script is too long. Maximum {MAX_SCRIPT_LENGTH} characters allowed. Your script has {len(script_text)} characters.") | |
| for i, line_content in enumerate(lines): | |
| line_content = line_content.strip() | |
| if not line_content: | |
| continue | |
| match = re.match(r'\[(.*?)\]\s*(.*)', line_content) | |
| if match: | |
| speaker, utterance = match.groups() | |
| utterance = utterance.strip() | |
| else: | |
| # If no speaker tag, assign a default speaker or handle as per requirements | |
| # For now, let's assume the whole line is an utterance by a "Narrator" or similar | |
| speaker = "Narrator" # Or consider raising an error/warning | |
| utterance = line_content.strip() | |
| if not utterance: # Skip if utterance is empty after parsing | |
| continue | |
| parsed_lines.append({"id": i, "speaker": speaker.strip(), "text": utterance}) | |
| total_chars += len(utterance) | |
| return parsed_lines, total_chars | |
| def calculate_cost(total_chars, num_lines, model_name="tts-1-hd"): | |
| """ | |
| Calculates the estimated cost for TTS processing. | |
| """ | |
| if model_name == "tts-1-hd": | |
| cost = total_chars * TTS_1_HD_COST_PER_CHAR | |
| elif model_name == "gpt-4o-mini-tts": | |
| # Estimate duration: total_chars / X chars per second | |
| # This is a very rough estimate. Actual duration depends on OpenAI's model. | |
| estimated_seconds = total_chars / CHARS_PER_SECOND_ESTIMATE | |
| cost = estimated_seconds * GPT_4O_MINI_TTS_COST_PER_SECOND | |
| else: | |
| raise ValueError(f"Unknown model for cost calculation: {model_name}") | |
| return cost | |
| if __name__ == '__main__': | |
| sample_script = """ | |
| [Alice] Hello Bob, how are you? | |
| [Bob] I'm fine, Alice. And you? | |
| This is a line without a speaker tag. | |
| [Charlie] Just listening in. | |
| """ | |
| parsed, chars = parse_dialogue_script(sample_script) | |
| print("Parsed Lines:") | |
| for p_line in parsed: | |
| print(p_line) | |
| print(f"\nTotal Characters: {chars}") | |
| cost_hd = calculate_cost(chars, len(parsed), "tts-1-hd") | |
| print(f"Estimated cost for tts-1-hd: ${cost_hd:.6f}") | |
| cost_gpt_mini = calculate_cost(chars, len(parsed), "gpt-4o-mini-tts") | |
| print(f"Estimated cost for gpt-4o-mini-tts: ${cost_gpt_mini:.6f}") | |
| long_script = "a" * (MAX_SCRIPT_LENGTH + 1) | |
| try: | |
| parse_dialogue_script(long_script) | |
| except ValueError as e: | |
| print(f"Error for long script: {e}") |