From 76fe99d3c9cae69a8666938fd72b7caf32d97583 Mon Sep 17 00:00:00 2001 From: arcelibs Date: Thu, 28 Dec 2023 07:50:58 +0800 Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E6=94=B9--NLP=E4=B8=80=E5=BE=8B?= =?UTF-8?q?=E4=BD=BF=E7=94=A8=20=E4=BF=AE=E6=94=B9--=E5=85=88=E4=BD=BF?= =?UTF-8?q?=E7=94=A8DeepL=20API=E5=86=8D=E4=BD=BF=E7=94=A8Gemini=20API=20?= =?UTF-8?q?=E4=BF=AE=E6=94=B9--blockReason:=20SAFETY=E6=99=82=E7=9B=B4?= =?UTF-8?q?=E6=8E=A5=E8=AA=BF=E7=94=A8DeepL=E7=B5=90=E6=9E=9C?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Local-Windows/main.py | 90 +++++++++++++++++++------------------------ 1 file changed, 39 insertions(+), 51 deletions(-) diff --git a/Local-Windows/main.py b/Local-Windows/main.py index 8cb201f..1a37e11 100644 --- a/Local-Windows/main.py +++ b/Local-Windows/main.py @@ -11,7 +11,7 @@ # 初始化 Faster Whisper 模型 -model = WhisperModel("large-v2", device="cuda", compute_type="int8") # 使用适合您硬件的模型大小和设备 +model = WhisperModel("large-v2", device="cuda", compute_type="int8") #從api_key.txt取得api def get_api_key_from_file(file_path='api_key.txt'): @@ -55,58 +55,49 @@ def call_deepl_api(input_text, deepL_auth_key, target_lang='ZH'): # 呼叫Gemini的函數 def call_gemini_api(input_text): - # 從文件中取得API KEY + # 获取 DeepL API KEY + deepl_auth_key = get_deepl_api_key_from_file() + if not deepl_auth_key: + print("无法获取 DeepL API KEY。") + return None + + # 先使用 DeepL API 翻译为简体中文 + simplified_chinese_text = call_deepl_api(input_text, deepl_auth_key, target_lang='ZH') + if not simplified_chinese_text: + print("DeepL 翻译失败。") + return None + + # 从文件中取得 Gemini API KEY api_key = get_api_key_from_file() if not api_key: - print("API KEY 無效") + print("Gemini API KEY 无效") return None - - # APIURL請求標頭 + + # Gemini API URL 请求头 url = "https://palm-proxy.arcelibs.com/v1beta/models/gemini-pro:generateContent?key={}".format(api_key) headers = {'Content-Type': 'application/json'} - # 分段處理 - segments = detect_and_split_text(input_text, max_length=500) # 假设 500 是分段长度的上限 - all_translated_text = [] + # 格式化输入文本，准备发送给 Gemini API + formatted_input = f"請你將下列語句翻譯成繁體中文，忽略任何語句與文法問題:\n{simplified_chinese_text}" + data = json.dumps({"contents": [{"parts": [{"text": formatted_input}]}]}) - for segment in segments: - formatted_input = f"請你必須將下列語句翻譯成流暢的繁體中文: \n{segment}" - data = json.dumps({"contents": [{"parts": [{"text": formatted_input}]}]}) + # 发送请求到 Gemini API + response = requests.post(url, headers=headers, data=data) + response_data = response.json() - response = requests.post(url, headers=headers, data=data) - response_data = response.json() - response_text = json.dumps(response_data, indent=4) - - # 保存API回應內容 - save_transcription("gemini_response", response_text, is_api_response=True) - - if response.status_code == 200: - if 'blockReason' in response_data and response_data['blockReason'] == 'SAFETY': - # 获取 DeepL API KEY - deepl_auth_key = get_deepl_api_key_from_file() - if deepl_auth_key: - deepl_translated = call_deepl_api(segment, deepl_auth_key) - if deepl_translated: - # 再次尝试使用 Gemini API - gemini_second_try = call_gemini_api(deepl_translated) - all_translated_text.append(gemini_second_try) - else: - print("DeepL 翻译失败。") - all_translated_text.append(f"[翻译段落失败: {segment}]") - else: - print("无法获取 DeepL API KEY。") - all_translated_text.append(f"[翻译段落失败: {segment}]") - elif 'candidates' in response_data: - translated_text = response_data['candidates'][0]['content']['parts'][0]['text'] - all_translated_text.append(translated_text) - else: - print("KeyError: 'candidates' not found in response.") - all_translated_text.append(f"[翻译段落失败: {segment}]") + # 检查响应并处理 + if response.status_code == 200: + if 'blockReason' in response_data and response_data['blockReason'] == 'SAFETY': + # 如果 Gemini API 返回 blockReason 为 SAFETY，直接使用 DeepL 翻译结果 + return simplified_chinese_text + elif 'candidates' in response_data: + return response_data['candidates'][0]['content']['parts'][0]['text'] else: - print(f"错误: {response.status_code}") - all_translated_text.append(f"[翻译段落失败: {segment}]") - - return ' '.join(all_translated_text) + print("KeyError: 'candidates' not found in response.") + return None + else: + print(f"错误: {response.status_code}") + return None # 使用 yt-dlp 獲取 YouTube 直播媒體位置 @@ -161,7 +152,7 @@ def save_transcription(file_path, text, is_api_response=False): file.write(text) # 新增語言檢測機制並加載模型 -def detect_and_split_text(text, max_length): +def detect_and_split_text(text, max_length=500): language = detect(text) nlp = None if language == "en": @@ -171,7 +162,7 @@ def detect_and_split_text(text, max_length): elif language in ["zh-cn", "zh-tw"]: nlp = spacy.load("zh_core_web_sm") else: - # 無法辨識?就用標點符號分段 + # 如果无法识别语言，则根据标点符号分段 return split_text_on_punctuation(text, max_length) return split_text_natural(nlp, text, max_length) @@ -185,7 +176,7 @@ def split_text_natural(nlp, text, max_length): if len(current_segment + sent.text) <= max_length: current_segment += sent.text + " " else: - if current_segment: # 空字串處理 + if current_segment: # 非空字符串 segments.append(current_segment.strip()) current_segment = sent.text + " " if current_segment: @@ -194,10 +185,8 @@ def split_text_natural(nlp, text, max_length): # 不是中英日文? 就用標點符號來分類 def split_text_on_punctuation(text, max_length): - # 定義分類的標點符號 punctuations = ".!?\n" pattern = f"[{re.escape(punctuations)}]" - segments = [] current_segment = "" for word in re.split(pattern, text): @@ -209,7 +198,6 @@ def split_text_on_punctuation(text, max_length): current_segment = word if current_segment: segments.append(current_segment.strip()) - return segments # 主流程 @@ -239,7 +227,7 @@ def main(segment_duration, total_duration): time.sleep(segment_duration) # 配置參數 -SEGMENT_DURATION = 10 # 每一段錄製的長度，單位是秒 +SEGMENT_DURATION = 15 # 每一段錄製的長度，單位是秒 TOTAL_DURATION = 6000 # 總錄製時間，單位是秒 if __name__ == "__main__":