File size: 4,119 Bytes
5c2ba64
2e2148b
5c2ba64
 
a9922ff
5c2ba64
 
 
 
672cb3f
a9922ff
 
5c2ba64
 
 
 
 
 
 
 
 
 
672cb3f
5c2ba64
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a9922ff
5c2ba64
a9922ff
5c2ba64
 
 
 
 
 
 
2e2148b
 
 
 
 
672cb3f
 
 
 
 
 
 
 
 
 
 
5c2ba64
2e2148b
672cb3f
f9493cb
 
 
672cb3f
 
5c2ba64
 
 
 
 
 
a9922ff
5c2ba64
 
 
 
2e2148b
5c2ba64
 
 
 
 
 
 
 
 
2e2148b
 
5c2ba64
 
672cb3f
 
 
2e2148b
 
 
 
 
5c2ba64
 
 
 
 
 
2e2148b
 
 
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
import io
import os
from typing import List, Literal

import gradio as gr
from loguru import logger
from openai import OpenAI
from promptic import llm
from pydantic import BaseModel
from pypdf import PdfReader


class DialogueItem(BaseModel):
    text: str
    voice: Literal["alloy", "onyx", "fable"]


class Dialogue(BaseModel):
    scratchpad: str
    dialogue: List[DialogueItem]


@llm(model="gemini/gemini-1.5-flash")
def generate_dialogue(text: str) -> Dialogue:
    """
    Your task is to take the input text provided and turn it into an engaging, informative podcast dialogue. The input text may be messy or unstructured, as it could come from a variety of sources like PDFs or web pages.

    Here is the input text you will be working with:

    ```
    {text}
    ```

    First, carefully read through the input text and identify the main topics, key points, and any interesting facts or anecdotes. Think about how you could present this information in a fun, engaging way that would be suitable for an audio podcast.

    Brainstorm creative ways to discuss the main topics and key points you identified in the input text. Consider using analogies, storytelling techniques, or hypothetical scenarios to make the content more relatable and engaging for listeners.

    Keep in mind that your podcast should be accessible to a general audience, so avoid using too much jargon or assuming prior knowledge of the topic. If necessary, think of ways to briefly explain any complex concepts in simple terms.

    Use your imagination to fill in any gaps in the input text or to come up with thought-provoking questions that could be explored in the podcast. The goal is to create an informative and entertaining dialogue, so feel free to be creative in your approach.

    Write your brainstorming ideas and a rough outline for the podcast dialogue in a scratchpad.

    Now that you have brainstormed ideas and created a rough outline, it's time to write the actual podcast dialogue. Aim for a natural, conversational flow between the host and any guest speakers. Incorporate the best ideas from your brainstorming session and make sure to explain any complex topics in an easy-to-understand way.

    Write your engaging, informative podcast dialogue based on the key points and creative ideas you came up with during the brainstorming session. Use a conversational tone and include any necessary context or explanations to make the content accessible to a general audience.
    """


def get_mp3(text: str, voice: str, api_key: str = None) -> bytes:
    client = OpenAI(
        api_key=api_key or os.getenv("OPENAI_API_KEY"),
    )

    with client.audio.speech.with_streaming_response.create(
        model="tts-1",
        voice=voice,
        input=text,
    ) as response:
        with io.BytesIO() as file:
            for chunk in response.iter_bytes():
                file.write(chunk)
            return file.getvalue()


def generate_audio(file: bytes, openai_api_key: str) -> bytes:

    if not os.getenv("OPENAI_API_KEY", openai_api_key):
        raise gr.Error("OpenAI API key is required")

    reader = PdfReader(io.BytesIO(file))
    text = "\n\n".join([page.extract_text() for page in reader.pages])

    llm_output = generate_dialogue(text)
    logger.info(llm_output)

    result = b""
    characters = 0

    for line in llm_output.dialogue:
        logger.info(line.text)
        logger.info(line.voice)

        audio = get_mp3(line.text, line.voice, openai_api_key)
        result += audio
        characters += len(line.text)

    logger.info(f"Generated {characters} characters of audio")

    return result


demo = gr.Interface(
    title="PDF to Podcast",
    description="Convert any PDF document into an engaging podcast episode!",
    fn=generate_audio,
    inputs=[
        gr.File(
            label="Input PDF",
            type="binary",
        ),
        gr.Textbox(
            label="OpenAI API Key",
            placeholder="Enter your OpenAI API key here",
        ),
    ],
    outputs=[
        gr.Audio(format="mp3"),
    ],
)

demo.launch(
    show_api=False,
)