mirror of
https://github.com/linyqh/NarratoAI.git
synced 2025-12-11 18:42:49 +00:00
优化剪辑逻辑
This commit is contained in:
linyq
parent
18d4fff028
commit
e75157a7ac
@ -126,7 +126,7 @@ def handle_exception(err):
|
||||
elif isinstance(err, ConnectionError):
|
||||
logger.error("网络连接错误，请检查您的网络连接。")
|
||||
else:
|
||||
logger.error(f"视频转录失败, 下面是具体报错信息: \n{traceback.format_exc()} \n问题排查指南: https://ai.google.dev/gemini-api/docs/troubleshooting?hl=zh-cn")
|
||||
logger.error(f"大模型请求失败, 下面是具体报错信息: \n{traceback.format_exc()}")
|
||||
return ""
|
||||
|
||||
|
||||
@ -353,11 +353,11 @@ def _generate_response(prompt: str, llm_provider: str = None) -> str:
|
||||
return content.replace("\n", "")
|
||||
|
||||
|
||||
def _generate_response_video(prompt: str, llm_provider: str, video_file: str | File) -> str:
|
||||
def _generate_response_video(prompt: str, llm_provider_video: str, video_file: str | File) -> str:
|
||||
"""
|
||||
多模态能力大模型
|
||||
"""
|
||||
if llm_provider == "gemini":
|
||||
if llm_provider_video == "gemini":
|
||||
api_key = config.app.get("gemini_api_key")
|
||||
model_name = config.app.get("gemini_model_name")
|
||||
base_url = "***"
|
||||
@ -366,7 +366,7 @@ def _generate_response_video(prompt: str, llm_provider: str, video_file: str | F
|
||||
"llm_provider 未设置，请在 config.toml 文件中进行设置。"
|
||||
)
|
||||
|
||||
if llm_provider == "gemini":
|
||||
if llm_provider_video == "gemini":
|
||||
import google.generativeai as genai
|
||||
|
||||
genai.configure(api_key=api_key, transport="rest")
|
||||
@ -441,15 +441,44 @@ def generate_script(
|
||||
compressed_video_path = f"{os.path.splitext(video_path)[0]}_compressed.mp4"
|
||||
compress_video(video_path, compressed_video_path)
|
||||
|
||||
# 2. 转录视频
|
||||
transcription = gemini_video_transcription(
|
||||
video_name=video_name,
|
||||
video_path=compressed_video_path,
|
||||
language=language,
|
||||
progress_text=progress_text,
|
||||
llm_provider="gemini"
|
||||
)
|
||||
|
||||
# # 2. 转录视频
|
||||
# transcription = gemini_video_transcription(
|
||||
# video_name=video_name,
|
||||
# video_path=compressed_video_path,
|
||||
# language=language,
|
||||
# progress_text=progress_text,
|
||||
# llm_provider_video="gemini"
|
||||
# )
|
||||
transcription = """
|
||||
[{"timestamp": "00:00-00:06", "picture": "一个穿着蓝色囚服，戴着手铐的人在房间里走路。", "speech": ""},
|
||||
{"timestamp": "00:06-00:09", "picture": "一个穿着蓝色囚服，戴着手铐的人，画面上方显示“李自忠 银行抢劫犯”。", "speech": "李自忠 银行抢劫一案 现在宣判"},
|
||||
{"timestamp": "00:09-00:12", "picture": "一个穿着黑色西装，打着红色领带的女人，坐在一个牌子上，牌子上写着“书记员”，身后墙上挂着“国徽”。", "speech": "全体起立"},
|
||||
{"timestamp": "00:12-00:15", "picture": "一个穿着黑色法官服的男人坐在一个牌子后面，牌子上写着“审判长”，身后墙上挂着“国徽”。法庭上，很多人站着。", "speech": ""},
|
||||
{"timestamp": "00:15-00:19", "picture": "一个穿着黑色西装，打着红色领带的女人，坐在一个牌子上，牌子上写着“书记员”，身后墙上挂着“国徽”。法庭上，很多人站着。", "speech": "本庭二审判决如下 被告李自忠 犯抢劫银行罪"},
|
||||
{"timestamp": "00:19-00:24", "picture": "一个穿着蓝色囚服，戴着手铐的人，画面上方显示“李自忠 银行抢劫犯”。", "speech": "维持一审判决 判处有期徒刑 二十年"},
|
||||
{"timestamp": "00:24-00:27", "picture": "一个穿着黑色法官服的男人坐在一个牌子后面，牌子上写着“审判长”，他敲了一下法槌。", "speech": ""},
|
||||
{"timestamp": "00:27-00:32", "picture": "一个穿着蓝色囚服，戴着手铐的人，画面上方显示“李自忠 银行抢劫犯”。", "speech": "我们要让她们牢底坐穿 越父啊越父 你一个平头老百姓 也敢跟外资银行做对 真是不知天高地厚"},
|
||||
{"timestamp": "00:32-00:41", "picture": "一个穿着蓝色囚服，戴着手铐的人跪在地上。", "speech": "我要让她们牢底坐穿 越父啊越父 你一个平头老百姓 也敢跟外资银行做对 真是不知天高地厚"},
|
||||
{"timestamp": "00:41-00:47", "picture": "两个警察押解着一个穿着蓝色囚服，戴着手铐的人走在路上，一个女记者在路边报道新闻。", "speech": "李先生 这里是孔雀卫视 这里是黄金眼819新闻直播间 这里是浙江卫视新闻直播间 近日李自忠案引发社会热议"},
|
||||
{"timestamp": "00:47-01:03", "picture": "一个穿着灰色外套的男人坐在银行柜台前，和银行工作人员说话。画面中还穿插着女记者在路边报道新闻的画面。", "speech": "李自忠案引发社会热议 李自忠在去银行取钱的时候 由于他拿的是儿子的存折 所以银行要求李自忠证明他的儿子就是他的儿子 我说取不了就是取不了啊 这是你儿子的存折啊 你要证明你儿子是你儿子啊"},
|
||||
{"timestamp": "01:03-01:10", "picture": "一个穿着灰色外套的男人坐在银行柜台前，和银行工作人员说话。画面中还穿插着女记者在路边报道新闻的画面。", "speech": "李自忠提供了身份证账户户口本后 银行都不认可他的儿子是他的儿子 就在这个时候 银行发生一起抢劫案"},
|
||||
{"timestamp": "01:10-01:17", "picture": "三个戴着帽子和口罩的劫匪持枪闯入银行，银行里的人都很害怕，纷纷蹲下躲避。", "speech": "都给我蹲下 老实点 把钱给我交出来"},
|
||||
{"timestamp": "01:17-01:28", "picture": "女记者在路边报道新闻，画面中穿插着银行抢劫案的画面。", "speech": "劫匪看到一旁大哭的李自忠 得知他是因为儿子需要治病才取钱的时候 给了他一打钱 怎么 你儿子在医院等着钱救命啊 银行不给取啊"},
|
||||
{"timestamp": "01:28-01:36", "picture": "一个戴着黑色帽子和口罩的劫匪，拿着枪，给一个穿着灰色外套的男人一叠钱。", "speech": "银行不给取啊 好了 给儿子看病去 李自忠在把钱给儿子交完药费后被捕"},
|
||||
{"timestamp": "01:36-01:58", "picture": "两个警察押解着一个穿着蓝色囚服，戴着手铐的男人走在路上，一个女记者在路边报道新闻。", "speech": "目前一审二审都维持原判 判处有期徒刑二十年 对此你有什么想说的吗 他怎么证明他儿子是他儿子 要是银行早点把钱给我 我也不会遇到劫匪 我儿子还得救命 不是的 儿子 儿子 儿子"},
|
||||
{"timestamp": "01:58-02:03", "picture": "两个警察押解着一个穿着蓝色囚服，戴着手铐的男人走在路上，一个女记者在路边报道新闻。男人情绪激动，大声喊叫。", "speech": "儿子 儿子 儿子"},
|
||||
{"timestamp": "02:03-02:12", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边。画面中穿插着新闻报道的画面。", "speech": "近日李自忠案引发社会热议 李自忠在去银行取钱的时候 银行要求李自忠证明他的儿子就是他的儿子"},
|
||||
{"timestamp": "02:12-02:25", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生站在门口。", "speech": "爸 这家人也真够可怜的 当爹的坐牢 这儿子 恐怕要成植物人了"},
|
||||
{"timestamp": "02:25-02:31", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生站在门口。", "speech": "医生啊 我弟弟的情况怎么样 我先看看"},
|
||||
{"timestamp": "02:31-02:40", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。", "speech": ""},
|
||||
{"timestamp": "02:40-02:46", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。", "speech": "不太理想啊 你弟弟想要醒过来 希望渺茫"},
|
||||
{"timestamp": "02:46-02:57", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。", "speech": "这 麟木 麟木你别吓姐啊麟木 麟木"},
|
||||
{"timestamp": "02:57-03:02", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。画面中穿插着新闻报道的画面。", "speech": "麟木 儿子 麟木你别吓姐啊麟木"},
|
||||
{"timestamp": "03:02-03:08", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。画面中穿插着新闻报道的画面。女人情绪激动，大声哭泣。", "speech": "儿子 麟木你别吓姐啊麟木 儿子"},
|
||||
{"timestamp": "03:08-03:14", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，一个穿着粉色上衣的女人站在病床边，一个白头发的医生正在给男人做检查。画面中穿插着新闻报道的画面。女人情绪激动，大声哭泣。", "speech": "儿子"},
|
||||
{"timestamp": "03:14-03:18", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，画面变成紫色光效。", "speech": ""},
|
||||
{"timestamp": "03:18-03:20", "picture": "一个病房里，一个年轻男人躺在病床上，戴着呼吸机，他突然睁开了眼睛。", "speech": ""}]
|
||||
"""
|
||||
# # 清理压缩后的视频文件
|
||||
# try:
|
||||
# os.remove(compressed_video_path)
|
||||
@ -458,7 +487,7 @@ def generate_script(
|
||||
|
||||
# 3. 编写解说文案
|
||||
progress_text.text("解说文案中...")
|
||||
script = writing_short_play(video_plot, video_name, "openai")
|
||||
script = writing_short_play(video_plot, video_name, "openai", count=300)
|
||||
|
||||
# 4. 文案匹配画面
|
||||
if transcription != "":
|
||||
@ -620,7 +649,7 @@ def gemini_video2json(video_origin_name: str, video_origin_path: str, video_plot
|
||||
return response
|
||||
|
||||
|
||||
def gemini_video_transcription(video_name: str, video_path: str, language: str, llm_provider: str, progress_text: st.empty = ""):
|
||||
def gemini_video_transcription(video_name: str, video_path: str, language: str, llm_provider_video: str, progress_text: st.empty = ""):
|
||||
'''
|
||||
使用 gemini-1.5-xxx 进行视频画面转录
|
||||
'''
|
||||
@ -661,28 +690,25 @@ def gemini_video_transcription(video_name: str, video_path: str, language: str,
|
||||
|
||||
progress_text.text("视频转录中...")
|
||||
try:
|
||||
response = _generate_response_video(prompt=prompt, llm_provider=llm_provider, video_file=gemini_video_file)
|
||||
response = _generate_response_video(prompt=prompt, llm_provider_video=llm_provider_video, video_file=gemini_video_file)
|
||||
logger.success("视频转录成功")
|
||||
logger.debug(response)
|
||||
print(type(response))
|
||||
return response
|
||||
except Exception as err:
|
||||
return handle_exception(err)
|
||||
|
||||
|
||||
def writing_movie(video_plot, video_name):
|
||||
def writing_movie(video_plot, video_name, llm_provider):
|
||||
"""
|
||||
影视解说（电影解说）
|
||||
"""
|
||||
api_key = config.app.get("gemini_api_key")
|
||||
model_name = config.app.get("gemini_model_name")
|
||||
|
||||
gemini.configure(api_key=api_key)
|
||||
model = gemini.GenerativeModel(model_name)
|
||||
|
||||
prompt = f"""
|
||||
**角色设定：**
|
||||
你是一名有10年经验的影视解说文案的创作者，
|
||||
下面是关于如何写解说文案的方法 {Method}，请认真阅读它，之后我会给你一部影视作品的名称，然后让你写一篇文案
|
||||
请根据方法撰写 《{video_name}》的影视解说文案，文案要符合以下要求:
|
||||
请根据方法撰写 《{video_name}》的影视解说文案，《{video_name}》的大致剧情如下: {video_plot}
|
||||
文案要符合以下要求:
|
||||
|
||||
**任务目标：**
|
||||
1. 文案字数在 1500字左右，严格要求字数，最低不得少于 1000字。
|
||||
@ -691,34 +717,17 @@ def writing_movie(video_plot, video_name):
|
||||
4. 不要包含小标题，每个段落以 \n 进行分隔。
|
||||
"""
|
||||
try:
|
||||
response = model.generate_content(
|
||||
prompt,
|
||||
generation_config=gemini.types.GenerationConfig(
|
||||
candidate_count=1,
|
||||
temperature=1.3,
|
||||
),
|
||||
safety_settings={
|
||||
HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE,
|
||||
HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
|
||||
HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
|
||||
HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
|
||||
}
|
||||
)
|
||||
return response.text
|
||||
response = _generate_response(prompt, llm_provider)
|
||||
logger.success("解说文案生成成功")
|
||||
return response
|
||||
except Exception as err:
|
||||
return handle_exception(err)
|
||||
|
||||
|
||||
def writing_short_play(video_plot: str, video_name: str, llm_provider: str):
|
||||
def writing_short_play(video_plot: str, video_name: str, llm_provider: str, count: int = 500):
|
||||
"""
|
||||
影视解说（短剧解说）
|
||||
"""
|
||||
# api_key = config.app.get("gemini_api_key")
|
||||
# # model_name = config.app.get("gemini_model_name")
|
||||
#
|
||||
# gemini.configure(api_key=api_key)
|
||||
# model = gemini.GenerativeModel(model_name)
|
||||
|
||||
if not video_plot:
|
||||
raise ValueError("短剧的简介不能为空")
|
||||
if not video_name:
|
||||
@ -732,27 +741,12 @@ def writing_short_play(video_plot: str, video_name: str, llm_provider: str):
|
||||
文案要符合以下要求:
|
||||
|
||||
**任务目标：**
|
||||
1. 文案字数在 800字左右，严格要求字数，最低不得少于 600字。
|
||||
1. 请严格要求文案字数, 字数控制在 {count} 字左右。
|
||||
2. 避免使用 markdown 格式输出文案。
|
||||
3. 仅输出解说文案，不输出任何其他内容。
|
||||
4. 不要包含小标题，每个段落以 \\n 进行分隔。
|
||||
"""
|
||||
try:
|
||||
# if "gemini" in model_name:
|
||||
# response = model.generate_content(
|
||||
# prompt,
|
||||
# generation_config=gemini.types.GenerationConfig(
|
||||
# candidate_count=1,
|
||||
# temperature=1.0,
|
||||
# ),
|
||||
# safety_settings={
|
||||
# HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE,
|
||||
# HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
|
||||
# HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
|
||||
# HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
|
||||
# }
|
||||
# )
|
||||
# else:
|
||||
response = _generate_response(prompt, llm_provider)
|
||||
logger.success("解说文案生成成功")
|
||||
logger.debug(response)
|
||||
@ -763,7 +757,7 @@ def writing_short_play(video_plot: str, video_name: str, llm_provider: str):
|
||||
|
||||
def screen_matching(huamian: str, wenan: str, llm_provider: str):
|
||||
"""
|
||||
画面匹配
|
||||
画面匹配（一次性匹配）
|
||||
"""
|
||||
if not huamian:
|
||||
raise ValueError("画面不能为空")
|
||||
@ -772,12 +766,18 @@ def screen_matching(huamian: str, wenan: str, llm_provider: str):
|
||||
|
||||
prompt = """
|
||||
你是一名有10年经验的影视解说创作者，
|
||||
你的任务是根据画面描述文本和解说文案，匹配出每段解说文案对应的画面时间戳, 结果以 json 格式输出。
|
||||
你的任务是根据视频转录脚本和解说文案，匹配出每段解说文案对应的画面时间戳, 结果以 json 格式输出。
|
||||
|
||||
画面描述文本和文案（由 XML 标记<SOURCE_TEXT><SOURCE_TEXT>和 <COPYWRITER><COPYWRITER>分隔）如下所示：
|
||||
<SOURCE_TEXT>
|
||||
注意：
|
||||
转录脚本中
|
||||
- timestamp: 表示视频时间戳
|
||||
- picture: 表示当前画面描述
|
||||
- speech": 表示当前视频中人物的台词
|
||||
|
||||
转录脚本和文案（由 XML 标记<PICTURE></PICTURE>和 <COPYWRITER></COPYWRITER>分隔）如下所示：
|
||||
<PICTURE>
|
||||
%s
|
||||
</SOURCE_TEXT>
|
||||
</PICTURE>
|
||||
|
||||
<COPYWRITER>
|
||||
%s
|
||||
@ -787,8 +787,15 @@ def screen_matching(huamian: str, wenan: str, llm_provider: str):
|
||||
- 使用以下 JSON schema:
|
||||
script = {'picture': str, 'timestamp': str(时间戳), "narration": str, "OST": bool(是否开启原声)}
|
||||
Return: list[script]
|
||||
- picture: 字段表示当前画面描述，与转录脚本保持一致
|
||||
- timestamp: 字段表示某一段文案对应的画面的时间戳，不必和转录脚本的时间戳一致，应该充分考虑文案内容，匹配出与其描述最匹配的时间戳
|
||||
- narration: 字段表示需要解说文案，每段解说文案尽量不要超过30字
|
||||
- OST: 字段表示是否开启原声，即当 OST 字段为 true 时，narration 字段为空字符串，当 OST 为 false 时，narration 字段为对应的解说文案
|
||||
- 注意，在画面匹配的过程中，需要适当的加入原声播放，使得解说和画面更加匹配，请按照 1:1 的比例，生成原声和解说的脚本内容。
|
||||
- 注意，在时间戳匹配上，一定不能原样照搬“转录脚本”，应当适当的合并或者删减一些片段。
|
||||
- 注意，第一个画面一定是原声播放并且时长不少于 20 s，为了吸引观众，第一段一定是整个转录脚本中最精彩的片段。
|
||||
- 注意，匹配的画面不能重复出现，即生成的脚本中 timestamp 不能重复。
|
||||
- 请以严格的 JSON 格式返回数据，不要包含任何注释、标记或其他字符。数据应符合 JSON 语法，可以被 json.loads() 函数直接解析， 不要添加 ```json 或其他标记。
|
||||
-
|
||||
""" % (huamian, wenan)
|
||||
try:
|
||||
response = _generate_response(prompt, llm_provider)
|
||||
|
||||
@ -1,4 +1,5 @@
|
||||
import os
|
||||
import subprocess
|
||||
import random
|
||||
from urllib.parse import urlencode
|
||||
|
||||
@ -329,6 +330,85 @@ def clip_videos(task_id: str, timestamp_terms: List[str], origin_video: str, ) -
|
||||
return video_paths
|
||||
|
||||
|
||||
def merge_videos(video_paths, ost_list):
|
||||
"""
|
||||
合并多个视频为一个视频，可选择是否保留每个视频的原声。
|
||||
|
||||
:param video_paths: 视频文件路径列表
|
||||
:param ost_list: 是否保留原声的布尔值列表
|
||||
:return: 合并后的视频文件路径
|
||||
"""
|
||||
if len(video_paths) != len(ost_list):
|
||||
raise ValueError("视频路径列表和保留原声列表长度必须相同")
|
||||
|
||||
if not video_paths:
|
||||
raise ValueError("视频路径列表不能为空")
|
||||
|
||||
# 准备临时文件列表
|
||||
temp_file = "temp_file_list.txt"
|
||||
with open(temp_file, "w") as f:
|
||||
for video_path, keep_ost in zip(video_paths, ost_list):
|
||||
if keep_ost:
|
||||
f.write(f"file '{video_path}'\n")
|
||||
else:
|
||||
# 如果不保留原声，创建一个无声的临时视频
|
||||
silent_video = f"silent_{os.path.basename(video_path)}"
|
||||
subprocess.run(["ffmpeg", "-i", video_path, "-c:v", "copy", "-an", silent_video], check=True)
|
||||
f.write(f"file '{silent_video}'\n")
|
||||
|
||||
# 合并视频
|
||||
output_file = "combined.mp4"
|
||||
ffmpeg_cmd = [
|
||||
"ffmpeg",
|
||||
"-f", "concat",
|
||||
"-safe", "0",
|
||||
"-i", temp_file,
|
||||
"-c:v", "copy",
|
||||
"-c:a", "aac",
|
||||
"-strict", "experimental",
|
||||
output_file
|
||||
]
|
||||
|
||||
try:
|
||||
subprocess.run(ffmpeg_cmd, check=True)
|
||||
print(f"视频合并成功：{output_file}")
|
||||
except subprocess.CalledProcessError as e:
|
||||
print(f"视频合并失败：{e}")
|
||||
return None
|
||||
finally:
|
||||
# 清理临时文件
|
||||
os.remove(temp_file)
|
||||
for video_path, keep_ost in zip(video_paths, ost_list):
|
||||
if not keep_ost:
|
||||
silent_video = f"silent_{os.path.basename(video_path)}"
|
||||
if os.path.exists(silent_video):
|
||||
os.remove(silent_video)
|
||||
|
||||
return output_file
|
||||
|
||||
|
||||
# 使用示例
|
||||
# if __name__ == "__main__":
|
||||
# video_paths = ['/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_17-01_37.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_00-00_06.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_06-00_09.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_03-01_10.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_10-01_17.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_24-00_27.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_28-01_36.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_32-00_41.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_36-01_58.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_12-00_15.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-00_09-00_12.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-02_12-02_25.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-02_03-02_12.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-01_58-02_03.mp4',
|
||||
# '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-03_14-03_18.mp4', '/Users/apple/Desktop/home/NarratoAI/storage/cache_videos/vid-03_18-03_20.mp4']
|
||||
#
|
||||
# ost_list = [True, False, False, False, False, False, False, False, True, False, False, False, False, False, False,
|
||||
# False]
|
||||
#
|
||||
# result = merge_videos(video_paths, ost_list)
|
||||
# if result:
|
||||
# print(f"合并后的视频文件：{result}")
|
||||
# else:
|
||||
# print("视频合并失败")
|
||||
#
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
download_videos(
|
||||
"test123", ["Money Exchange Medium"], audio_duration=100, source="pixabay"
|
||||
|
||||
@ -211,7 +211,7 @@ def start(task_id, params: VideoParams, stop_at: str = "video"):
|
||||
|
||||
if type(params.video_concat_mode) is str:
|
||||
params.video_concat_mode = VideoConcatMode(params.video_concat_mode)
|
||||
|
||||
|
||||
# 1. Generate script
|
||||
video_script = generate_script(task_id, params)
|
||||
if not video_script:
|
||||
@ -323,7 +323,7 @@ def start(task_id, params: VideoParams, stop_at: str = "video"):
|
||||
return kwargs
|
||||
|
||||
|
||||
def start_subclip(task_id, params: VideoClipParams, subclip_path_videos):
|
||||
def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos):
|
||||
"""
|
||||
后台任务（自动剪辑视频进行剪辑）
|
||||
|
||||
@ -423,39 +423,46 @@ def start_subclip(task_id, params: VideoClipParams, subclip_path_videos):
|
||||
combined_video_paths = []
|
||||
|
||||
_progress = 50
|
||||
for i in range(params.video_count):
|
||||
index = i + 1
|
||||
combined_video_path = path.join(utils.task_dir(task_id), f"combined-{index}.mp4")
|
||||
logger.info(f"\n\n## 5. 合并视频: {index} => {combined_video_path}")
|
||||
video.combine_clip_videos(
|
||||
combined_video_path=combined_video_path,
|
||||
video_paths=subclip_videos,
|
||||
video_ost_list=video_ost,
|
||||
list_script=list_script,
|
||||
video_aspect=params.video_aspect,
|
||||
threads=1 # 暂时只支持单线程
|
||||
)
|
||||
# for i in range(params.video_count):
|
||||
index = 1
|
||||
combined_video_path = path.join(utils.task_dir(task_id), f"combined.mp4")
|
||||
logger.info(f"\n\n## 5. 合并视频: => {combined_video_path}")
|
||||
print("111", subclip_videos)
|
||||
print("222", video_ost)
|
||||
print("333", len(subclip_videos))
|
||||
print("444", len(video_ost))
|
||||
# for video_path, video_ost in zip(subclip_videos, video_ost):
|
||||
# print(video_path)
|
||||
# print(video_ost)
|
||||
video.combine_clip_videos(
|
||||
combined_video_path=combined_video_path,
|
||||
video_paths=subclip_videos,
|
||||
video_ost_list=video_ost,
|
||||
list_script=list_script,
|
||||
video_aspect=params.video_aspect,
|
||||
threads=1 # 暂时只支持单线程
|
||||
)
|
||||
|
||||
_progress += 50 / params.video_count / 2
|
||||
sm.state.update_task(task_id, progress=_progress)
|
||||
_progress += 50 / 2
|
||||
sm.state.update_task(task_id, progress=_progress)
|
||||
|
||||
final_video_path = path.join(utils.task_dir(task_id), f"final-{index}.mp4")
|
||||
final_video_path = path.join(utils.task_dir(task_id), f"final-{index}.mp4")
|
||||
|
||||
logger.info(f"\n\n## 6. 最后一步: {index} => {final_video_path}")
|
||||
# 把所有东西合到在一起
|
||||
video.generate_video_v2(
|
||||
video_path=combined_video_path,
|
||||
audio_paths=audio_files,
|
||||
subtitle_path=subtitle_path,
|
||||
output_file=final_video_path,
|
||||
params=params,
|
||||
)
|
||||
logger.info(f"\n\n## 6. 最后一步: {index} => {final_video_path}")
|
||||
# 把所有东西合到在一起
|
||||
video.generate_video_v2(
|
||||
video_path=combined_video_path,
|
||||
audio_paths=audio_files,
|
||||
subtitle_path=subtitle_path,
|
||||
output_file=final_video_path,
|
||||
params=params,
|
||||
)
|
||||
|
||||
_progress += 50 / params.video_count / 2
|
||||
sm.state.update_task(task_id, progress=_progress)
|
||||
_progress += 50 / 2
|
||||
sm.state.update_task(task_id, progress=_progress)
|
||||
|
||||
final_video_paths.append(final_video_path)
|
||||
combined_video_paths.append(combined_video_path)
|
||||
final_video_paths.append(final_video_path)
|
||||
combined_video_paths.append(combined_video_path)
|
||||
|
||||
logger.success(f"任务 {task_id} 已完成, 生成 {len(final_video_paths)} 个视频.")
|
||||
|
||||
@ -468,11 +475,25 @@ def start_subclip(task_id, params: VideoClipParams, subclip_path_videos):
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
task_id = "task_id"
|
||||
params = VideoParams(
|
||||
video_subject="金钱的作用",
|
||||
voice_name="zh-CN-XiaoyiNeural-Female",
|
||||
voice_rate=1.0,
|
||||
|
||||
task_id = "test123456"
|
||||
subclip_path_videos = {'01:17-01:37': './storage/cache_videos/vid-01_17-01_37.mp4',
|
||||
'00:00-00:06': './storage/cache_videos/vid-00_00-00_06.mp4',
|
||||
'00:06-00:09': './storage/cache_videos/vid-00_06-00_09.mp4',
|
||||
'01:03-01:10': './storage/cache_videos/vid-01_03-01_10.mp4',
|
||||
'01:10-01:17': './storage/cache_videos/vid-01_10-01_17.mp4',
|
||||
'00:24-00:27': './storage/cache_videos/vid-00_24-00_27.mp4',
|
||||
'01:28-01:36': './storage/cache_videos/vid-01_28-01_36.mp4',
|
||||
'00:32-00:41': './storage/cache_videos/vid-00_32-00_41.mp4',
|
||||
'01:36-01:58': './storage/cache_videos/vid-01_36-01_58.mp4',
|
||||
'00:12-00:15': './storage/cache_videos/vid-00_12-00_15.mp4',
|
||||
'00:09-00:12': './storage/cache_videos/vid-00_09-00_12.mp4',
|
||||
'02:12-02:25': './storage/cache_videos/vid-02_12-02_25.mp4',
|
||||
'02:03-02:12': './storage/cache_videos/vid-02_03-02_12.mp4',
|
||||
'01:58-02:03': './storage/cache_videos/vid-01_58-02_03.mp4',
|
||||
'03:14-03:18': './storage/cache_videos/vid-03_14-03_18.mp4',
|
||||
'03:18-03:20': './storage/cache_videos/vid-03_18-03_20.mp4'}
|
||||
params = VideoClipParams(
|
||||
video_clip_json_path="/Users/apple/Desktop/home/NarratoAI/resource/scripts/test003.json",
|
||||
video_origin_path="/Users/apple/Desktop/home/NarratoAI/resource/videos/1.mp4",
|
||||
)
|
||||
start(task_id, params, stop_at="video")
|
||||
start_subclip(task_id, params, subclip_path_videos=subclip_path_videos)
|
||||
|
||||
@ -521,7 +521,8 @@ def combine_clip_videos(combined_video_path: str,
|
||||
Args:
|
||||
combined_video_path: 合并后的存储路径
|
||||
video_paths: 子视频路径列表
|
||||
audio_file: mp3旁白
|
||||
video_ost_list: 原声播放列表
|
||||
list_script: 剪辑脚本
|
||||
video_aspect: 屏幕比例
|
||||
threads: 线程数
|
||||
|
||||
@ -545,10 +546,13 @@ def combine_clip_videos(combined_video_path: str,
|
||||
# 一遍又一遍地添加下载的剪辑，直到达到音频的持续时间 （max_duration）
|
||||
# while video_duration < audio_duration:
|
||||
for video_path, video_ost in zip(video_paths, video_ost_list):
|
||||
clip = VideoFileClip(video_path)
|
||||
# 通过 ost 字段判断是否播放原声
|
||||
if not video_ost:
|
||||
clip = clip.without_audio()
|
||||
print(video_path)
|
||||
print(video_ost)
|
||||
cache_video_path = utils.root_dir()
|
||||
clip = VideoFileClip(os.path.join(cache_video_path, video_path))
|
||||
# # 通过 ost 字段判断是否播放原声
|
||||
# if not video_ost:
|
||||
# clip = clip.without_audio()
|
||||
# # 检查剪辑是否比剩余音频长
|
||||
# if (audio_duration - video_duration) < clip.duration:
|
||||
# clip = clip.subclip(0, (audio_duration - video_duration))
|
||||
|
||||
@ -1071,7 +1071,10 @@ def azure_tts_v1(
|
||||
(chunk["offset"], chunk["duration"]), chunk["text"]
|
||||
)
|
||||
return sub_maker
|
||||
|
||||
# 判断音频文件是否一件存在
|
||||
if os.path.exists(voice_file):
|
||||
logger.info(f"voice file exists, skip tts: {voice_file}")
|
||||
continue
|
||||
sub_maker = asyncio.run(_do())
|
||||
if not sub_maker or not sub_maker.subs:
|
||||
logger.warning(f"failed, sub_maker is None or sub_maker.subs is None")
|
||||
|
||||
@ -358,7 +358,7 @@ def add_new_timestamps(scenes):
|
||||
new_scene['new_timestamp'] = f"{new_start_str}-{new_end_str}"
|
||||
|
||||
# 为"原生播放"的narration添加唯一标识符
|
||||
if new_scene.get('narration') == "原声播放" or new_scene.get('narration') == None:
|
||||
if new_scene.get('narration') == "" or new_scene.get('narration') == None:
|
||||
unique_id = str(uuid4())[:8] # 使用UUID的前8个字符作为唯一标识符
|
||||
new_scene['narration'] = f"原声播放_{unique_id}"
|
||||
|
||||
|
||||
2
webui.py
2
webui.py
@ -456,7 +456,7 @@ with left_panel:
|
||||
video_script['path'] = subclip_videos[video_script['timestamp']]
|
||||
except KeyError as err:
|
||||
st.error(f"裁剪视频失败 {err}")
|
||||
# logger.debug(f"当前的脚本为：{st.session_state.video_script_list}")
|
||||
logger.debug(f"当前的脚本为：{st.session_state.subclip_videos}")
|
||||
else:
|
||||
st.error(tr("请先生成视频脚本"))
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user