File size: 5,598 Bytes
5fe16b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63ea22a
 
 
5fe16b1
 
 
 
 
 
 
63ea22a
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
from dataclasses import dataclass
from typing import List, Tuple, Dict
import os
import re
import httpx
import json
from openai import OpenAI
import edge_tts
import tempfile
import wave
from pydub import AudioSegment
import base64
from pathlib import Path


@dataclass
class ConversationConfig:
    max_words: int = 3000
    prefix_url: str = "https://r.jina.ai/"
    model_name: str = "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo"


class URLToAudioConverter:
    def __init__(self, config: ConversationConfig, llm_api_key: str):
        self.config = config
        self.llm_client = OpenAI(api_key=llm_api_key, base_url="https://api.together.xyz/v1")
        self.llm_out = None

    def fetch_text(self, url: str) -> str:
        if not url:
            raise ValueError("URL cannot be empty")

        full_url = f"{self.config.prefix_url}{url}"
        try:
            response = httpx.get(full_url, timeout=60.0)
            response.raise_for_status()
            return response.text
        except httpx.HTTPError as e:
            raise RuntimeError(f"Failed to fetch URL: {e}")

    def extract_conversation(self, text: str) -> Dict:
        if not text:
            raise ValueError("Input text cannot be empty")

        try:
            chat_completion = self.llm_client.chat.completions.create(
                messages=[{"role": "user", "content": self._build_prompt(text)}],
                model=self.config.model_name,
            )

            pattern = r"\{(?:[^{}]|(?:\{[^{}]*\}))*\}"
            json_match = re.search(pattern, chat_completion.choices[0].message.content)

            if not json_match:
                raise ValueError("No valid JSON found in response")

            return json.loads(json_match.group())
        except Exception as e:
            raise RuntimeError(f"Failed to extract conversation: {e}")

    def _build_prompt(self, text: str) -> str:
        template = """
        {
            "conversation": [
                {"speaker": "", "text": ""},
                {"speaker": "", "text": ""}
            ]
        }
        """
        return (
            f"{text}\nConvert the provided text into a short informative and crisp "
            f"podcast conversation between two experts. The tone should be "
            f"professional and engaging. Please adhere to the following "
            f"format and return the conversation in JSON:\n{template}"
        )

    async def text_to_speech(self, conversation_json: Dict, voice_1: str, voice_2: str) -> Tuple[List[str], str]:
        output_dir = Path(self._create_output_directory())
        filenames = []

        try:
            for i, turn in enumerate(conversation_json["conversation"]):
                filename = output_dir / f"output_{i}.wav"
                voice = voice_1 if i % 2 == 0 else voice_2

                tmp_path, error = await self._generate_audio(turn["text"], voice)
                if error:
                    raise RuntimeError(f"Text-to-speech failed: {error}")

                os.rename(tmp_path, filename)
                filenames.append(str(filename))

            return filenames, str(output_dir)
        except Exception as e:
            raise RuntimeError(f"Failed to convert text to speech: {e}")

    async def _generate_audio(self, text: str, voice: str, rate: int = 0, pitch: int = 0) -> Tuple[str, str]:
        if not text.strip():
            return None, "Text cannot be empty"
        if not voice:
            return None, "Voice cannot be empty"

        voice_short_name = voice.split(" - ")[0]
        rate_str = f"{rate:+d}%"
        pitch_str = f"{pitch:+d}Hz"
        communicate = edge_tts.Communicate(text, voice_short_name, rate=rate_str, pitch=pitch_str)

        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
            tmp_path = tmp_file.name
            await communicate.save(tmp_path)

        return tmp_path, None

    def _create_output_directory(self) -> str:
        random_bytes = os.urandom(8)
        folder_name = base64.urlsafe_b64encode(random_bytes).decode("utf-8")
        os.makedirs(folder_name, exist_ok=True)
        return folder_name

    def combine_audio_files(self, filenames: List[str], output_file: str) -> None:
        if not filenames:
            raise ValueError("No input files provided")

        try:
            audio_segments = []

            for filename in filenames:
                audio_segment = AudioSegment.from_mp3(filename)
                audio_segments.append(audio_segment)

            combined = sum(audio_segments)

            combined.export(output_file, format="wav")

            for filename in filenames:
                os.remove(filename)

        except Exception as e:
            raise RuntimeError(f"Failed to combine audio files: {e}")

    async def url_to_audio(self, url: str, voice_1: str, voice_2: str) -> str:
        text = self.fetch_text(url)

        words = text.split()
        if len(words) > self.config.max_words:
            text = " ".join(words[: self.config.max_words])

        conversation_json = self.extract_conversation(text)
        conversation_text = "\n".join(
        f"{turn['speaker']}: {turn['text']}" for turn in conversation_json["conversation"]
    )
        self.llm_out = conversation_json
        audio_files, folder_name = await self.text_to_speech(
            conversation_json, voice_1, voice_2
        )

        final_output = os.path.join(folder_name, "combined_output.wav")
        self.combine_audio_files(audio_files, final_output)
        return final_output,conversation_text