diff --git a/python/02-use-cases/ad_video_gen_seq/README.md b/python/02-use-cases/ad_video_gen_seq/README.md index 156fb45a..9e93e768 100644 --- a/python/02-use-cases/ad_video_gen_seq/README.md +++ b/python/02-use-cases/ad_video_gen_seq/README.md @@ -48,7 +48,7 @@ 开始前，请确保您的开发环境满足以下要求： - Python 3.12 或更高版本 -- veadk-python 0.5.5（见 `pyproject.toml`） +- veadk-python 0.5.22（见 `pyproject.toml`） - 推荐使用 `uv` 进行依赖管理 - 本地需要可用的 `ffmpeg`（用于 `moviepy` 合成视频） - 获取火山方舟 API KEY @@ -173,6 +173,7 @@ agentkit launch ## 示例提示词 以下是一些常用的提示词示例： + - `请生成一条巧克力的圣诞节营销视频。商品名：圣诞限定黑巧克力礼盒装。适用场景和人群：适合所有巧克力爱好者，特别是追求圣诞节极致口感、甜蜜分享与能量补充的消费者；适用于圣诞下午茶时光、节日亲友聚会、温馨赠礼或任何需要增添节日氛围的愉悦时刻。主要成分：精选可可豆、纯可可脂、优质牛奶、天然香草，无添加人工色素和防腐剂，富含抗氧化剂。口味/特点：入口即化，丝滑醇厚，浓郁可可香，微苦回甘中带有节日限定的暖心回味 http://lf3-static.bytednsdoc.com/obj/eden-cn/lm_sth/ljhwZthlaukjlkulzlp/ark/assistant/images/ad_chocolate.png` - `请生成一条面包营销视频。商品名：奶香松软拉丝吐司；适用场景和人群：场景：早餐配餐、下午茶点、日常代餐，人群：上班族、学生党、家庭群体（偏好松软口感的面包爱好者）；主要成分：高筋面粉、牛奶、鸡蛋、黄油、酵母、白砂糖；口味 / 特点：口味：浓郁奶香，搭配黄油 + 蜂蜜后口感香甜柔润，特点：面包质地松软、切面蜂窝气孔均匀，烤后表皮带焦脆斑点，兼具松软内里与香脆外皮的双重口感 http://lf3-static.bytednsdoc.com/obj/eden-cn/lm_sth/ljhwZthlaukjlkulzlp/ark/assistant/images/ad_bread.jpeg` - `生成一条侘寂风香薰蜡烛的商品图电商营销视频。商品名：侘寂风香薰蜡烛；适用场景和人群：场景：居家客厅装饰、卧室助眠、书房放松、极简风空间氛围营造，人群：喜欢极简 / 侘寂美学的家居爱好者、追求松弛感的都市白领、香薰收藏者；主要成分：天然大豆蜡、植物精油、水泥罐身、纸质标识贴；气味 / 特点：气味：可选木质调（雪松 / 檀香）、草本调（鼠尾草 / 尤加利）等中性舒缓香型，特点：水泥罐身带原生肌理质感，搭配黑白极简花纹标识，烛火柔和不刺眼；罐身可重复利用，整体风格低调质朴，契合侘寂美学的调性 http://lf3-static.bytednsdoc.com/obj/eden-cn/lm_sth/ljhwZthlaukjlkulzlp/ark/assistant/images/ad_candle.jpeg` diff --git a/python/02-use-cases/ad_video_gen_seq/README_en.md b/python/02-use-cases/ad_video_gen_seq/README_en.md index ccf6c10c..165d871a 100644 --- a/python/02-use-cases/ad_video_gen_seq/README_en.md +++ b/python/02-use-cases/ad_video_gen_seq/README_en.md @@ -48,7 +48,7 @@ The system exposes one Root Agent and orchestrates the following sub-agents in s Before starting, make sure your environment meets these requirements: - Python 3.12 or later -- veadk-python 0.5.5 (see `pyproject.toml`) +- veadk-python 0.5.22 (see `pyproject.toml`) - `uv` is recommended for dependency management - `ffmpeg` available locally (used by `moviepy` for video composition) - Get Volcengine Ark API KEY @@ -197,4 +197,3 @@ FAQ list to be added. ## License This project is open-sourced. See the LICENSE file in the repository root for details. - diff --git a/python/02-use-cases/ad_video_gen_seq/app/image/tools/image_generate_gather.py b/python/02-use-cases/ad_video_gen_seq/app/image/tools/image_generate_gather.py index c674f43d..a01cdfce 100644 --- a/python/02-use-cases/ad_video_gen_seq/app/image/tools/image_generate_gather.py +++ b/python/02-use-cases/ad_video_gen_seq/app/image/tools/image_generate_gather.py @@ -20,11 +20,19 @@ image_generate as image_generate_builtin, ) from veadk.utils.logger import get_logger +from veadk.config import getenv +from veadk.consts import DEFAULT_IMAGE_GENERATE_MODEL_NAME + logger = get_logger(__name__) -async def image_generate(tasks: list[dict], tool_context: ToolContext) -> Dict: +async def image_generate( + tasks: list[dict], + tool_context: ToolContext, + timeout: int = 600, + model_name: str = getenv("MODEL_IMAGE_NAME", DEFAULT_IMAGE_GENERATE_MODEL_NAME), +) -> Dict: """Generate images with Seedream 4.5. Commit batch image generation requests via tasks. @@ -32,6 +40,15 @@ async def image_generate(tasks: list[dict], tool_context: ToolContext) -> Dict: Args: tasks (list[dict]): A list of image-generation tasks. Each task is a dict. + timeout (int) + The timeout limit for the image generation task request, in seconds, with a default value of 600 seconds. + model_name (str): + Optional model name. If not specified, use the default model from environment. + If during execution, this tool encounters a model-related error (note that it must be a model-related error, otherwise do not perform this action), such as `ModelNotOpen`, + then after reminding about the relevant issue, you can execute this tool again and downgrade the model to the following models, passing this parameter: + - `doubao-seedream-5-0-260128` + - `doubao-seedream-4-5-251128` + - `doubao-seedream-4-0-250828` Per-task schema --------------- Required: @@ -89,7 +106,7 @@ async def image_generate(tasks: list[dict], tool_context: ToolContext) -> Dict: - 组图任务必须 sequential_image_generation="auto"。 - size 推荐使用 2048x2048 或表格里的标准比例，确保生成质量。 """ - logger.debug(f"image_generate_gather tasks: {tasks}") + logger.error(f"image_generate_gather tasks: {tasks}") new_tasks = [] task_origin_info = [] # Stores (original_task_index, sub_index_within_group) @@ -146,7 +163,9 @@ async def image_generate(tasks: list[dict], tool_context: ToolContext) -> Dict: # Call the underlying image_generate function with the flattened list of tasks logger.debug(f"image_generate_gather new_tasks: {new_tasks}") - raw_result = await image_generate_builtin(new_tasks, tool_context) + raw_result = await image_generate_builtin( + new_tasks, tool_context, timeout, model_name + ) logger.debug(f"image_generate_gather raw_result: {raw_result}") # Remap the results to match the original task structure @@ -166,6 +185,10 @@ async def image_generate(tasks: list[dict], tool_context: ToolContext) -> Dict: original_idx, original_sub_idx = task_origin_info[new_task_idx] new_key = f"task_{original_idx}_image_{original_sub_idx}" + logger.error( + f"image_generate_gather new_task_idx: {new_task_idx}, original_idx: {original_idx}, original_sub_idx: {original_sub_idx}" + ) + logger.error(f"image_generate_gather new_key: {new_key}, url: {url}") remapped_success.append({new_key: url}) for error_item in raw_result.get("error_list", []): diff --git a/python/02-use-cases/ad_video_gen_seq/app/video/tools/video_generate_by_code.py b/python/02-use-cases/ad_video_gen_seq/app/video/tools/video_generate_by_code.py index 337660d9..a8147c77 100644 --- a/python/02-use-cases/ad_video_gen_seq/app/video/tools/video_generate_by_code.py +++ b/python/02-use-cases/ad_video_gen_seq/app/video/tools/video_generate_by_code.py @@ -18,6 +18,8 @@ from veadk.tools.builtin_tools.video_generate import ( video_generate as video_generate_builtin, ) +from veadk.config import getenv +from veadk.consts import DEFAULT_VIDEO_MODEL_NAME async def video_generate( @@ -25,6 +27,7 @@ async def video_generate( tool_context: ToolContext, batch_size: int = 10, max_wait_seconds: int = 1200, + model_name: str = getenv("MODEL_VIDEO_NAME", DEFAULT_VIDEO_MODEL_NAME), ) -> Dict: """ Generate videos in **batch** from text prompts, optionally guided by a first/last frame, @@ -82,6 +85,13 @@ async def video_generate( - Size: < 30 MB - Base64 data URL example: `data:image/png;base64,` + Optional model_name(str): + If the user does not specify, do not pass this parameter and use the default value. + If during execution, this tool encounters a model-related error (note that it must be a model-related error, otherwise do not perform this action), such as `ModelNotOpen`, + then after reminding about the relevant issue, you can execute this tool again and downgrade the model to the following models, passing this parameter: + `doubao-seedance-1-5-pro-251215` + `doubao-seedance-1-0-pro-250528` + Model text commands (append after the prompt; unsupported keys are ignored by some models): --rs / --resolution Video resolution. Common values: 480p, 720p, 1080p. Default depends on model (e.g., doubao-seedance-1-0-pro: 1080p, @@ -148,5 +158,5 @@ async def video_generate( f"（可以有极其轻度的动作音，但禁止任何人声，禁止背景音乐，禁止音效，禁止旁白，禁止解说）{item['prompt']}" ) return await video_generate_builtin( - params, tool_context, batch_size, max_wait_seconds + params, tool_context, batch_size, max_wait_seconds, model_name ) diff --git a/python/02-use-cases/ad_video_gen_seq/pyproject.toml b/python/02-use-cases/ad_video_gen_seq/pyproject.toml index b50291d5..40ced70d 100644 --- a/python/02-use-cases/ad_video_gen_seq/pyproject.toml +++ b/python/02-use-cases/ad_video_gen_seq/pyproject.toml @@ -5,7 +5,7 @@ description = "Add your description here" readme = "README.md" requires-python = ">=3.12" dependencies = [ - "veadk-python==0.5.5", + "veadk-python==0.5.22", "uvicorn>=0.38.0", "moviepy>=2.2.1", "requests>=2.32.5", diff --git a/python/02-use-cases/ad_video_gen_seq/requirements.txt b/python/02-use-cases/ad_video_gen_seq/requirements.txt index 8647556e..73cf9f61 100644 --- a/python/02-use-cases/ad_video_gen_seq/requirements.txt +++ b/python/02-use-cases/ad_video_gen_seq/requirements.txt @@ -245,7 +245,7 @@ uritemplate==4.2.0 urllib3==2.6.2 uv==0.9.21 uvicorn==0.40.0 -veadk-python==0.5.5 +veadk-python==0.5.22 vikingdb-python-sdk==0.1.3 volcengine==1.0.212 volcengine-python-sdk==5.0.3