sanford

prompt = """I want you to create a storyboard based on the novel content, inferring scenes from the original text description; infer and supplement missing or implied information, including but not limited to: character clothing, character hairstyle, character hair color, character complexion, character facial features, character posture, character emotions, character body movements, etc.), style description (including but not limited to: era description, space description, time period description, geographical environment description, weather description), object description (including but not limited to: animals, plants, food, fruits, toys), visual perspective (including but not limited to: character proportions, camera depth description, observation angle description), but do not overdo it. Describe richer character emotions and emotional states through camera language, and after you understand this requirement, generate a new descriptive content through sentences. Change the output format to: Illustration 1: Original description: corresponding original sentences; Scene description: corresponding scene plot content; Scene characters: names of characters appearing in the scene; Clothing: protagonist in casual wear; Location: sitting in front of the bar; Expression: facial lines gentle, expression content; Behavior: gently shaking the wine glass in hand. Environment: the background of the bar is dark-toned, candlelight flickers in the background, giving a dreamy feeling. If you understand this requirement, please confirm these five points, and return results only with these five points' content, the novel content is as follows:"""

def split_text_into_chunks(text, max_length=ai_max_length):
    """
    Split text into chunks with a maximum length, ensuring that splits only occur at line breaks.
    """
    lines = text.splitlines()
    chunks = []
    current_chunk = ''
    for line in lines:
        if len(current_chunk + ' ' + line) <= max_length:
            current_chunk += ' ' + line
        else:
            chunks.append(current_chunk)
            current_chunk = line
    chunks.append(current_chunk)
    return chunks

def rewrite_text_with_genai(text, prompt="Please rewrite this text:"):
    chunks = split_text_into_chunks(text)
    rewritten_text = ''
    # pbar = tqdm(total=len(chunks), ncols=150)
    genai.configure(api_key=cfg['genai_api_key'])
    model = genai.GenerativeModel('gemini-pro')
    for chunk in chunks:
        _prompt=f"{prompt}\n{chunk}",
        response = model.generate_content(
            contents=_prompt, 
            generation_config=genai.GenerationConfig(
                temperature=0.1,
            ),
            stream=True,
            safety_settings = [
                {
                    "category": "HARM_CATEGORY_DANGEROUS",
                    "threshold": "BLOCK_NONE",
                },
                {
                    "category": "HARM_CATEGORY_HARASSMENT",
                    "threshold": "BLOCK_NONE",
                },
                {
                    "category": "HARM_CATEGORY_HATE_SPEECH",
                    "threshold": "BLOCK_NONE",
                },
                {
                    "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
                    "threshold": "BLOCK_NONE",
                },
                {
                    "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
                    "threshold": "BLOCK_NONE",
                },
            ]
        )
        for _chunk in response:
            if _chunk.text is not None:
                rewritten_text += _chunk.text.strip()
    #     pbar.update(1)
    # pbar.close()
    return rewritten_text

from diffusers import StableDiffusionPipeline
from diffusers.utils import load_image
import torch

model_path = "./models/cetusMix_Whalefall2.safetensors"
pipeline = StableDiffusionPipeline.from_single_file(
    model_path,
    torch_dtype=torch.float16,
    variant="fp16"
    ).to("mps")
generator = torch.Generator("mps").manual_seed(31)

def sd_cetus(save_name, prompt):
    prompt = prompt
    image = pipeline(prompt).images[0]
    image.save('data/img/'+ save_name +'.jpg')

import edge_tts
import asyncio

voice = 'zh-CN-YunxiNeural'
output = 'data/voice/'
rate = '-4%'
volume = '+0%'

async def tts_function(text, save_name):
    tts = edge_tts.Communicate(
        text,
        voice=voice,
        rate=rate,
        volume=volume
        )
    await tts.save(output + save_name + '.wav')