The video intelligent inpainting operator takes a single video as input and outputs a repaired video with specified content erased. Supports subtitles, scrolling subtitles, text watermarks, corner marks, logos, and episode information, and allows separate control of the erasure range for subtitles and watermarks.

Operator ID: las_video_inpaint_pro

Core capabilities #

• standard_subtitle: Standard erasure of subtitles and scrolling subtitles; default mode.
• standard_watermark: Standard erasure of text watermarks; best suited for static text watermarks in the corners of the frame.
• standard_subtitle_watermark: Standard simultaneous erasure of subtitles, scrolling subtitles, and text watermarks.
• precise: When non-empty fixed_bboxes are not provided, non-empty targets can be used to precisely specify watermarks, subtitles, scrolling subtitles, or episode information; when non-empty fixed_bboxes are provided, fixed regions are used directly and targets are not required. The standard mode automatically selects objects and ignores the targets parameter.
• Supports specifying one or more fixed inpainting regions using fixed_bboxes. Coordinates must satisfy top_left_x < bottom_right_x and top_left_y < bottom_right_y.
• Supports stable processing of long videos and retains the original audio in the output video.

Supported languages #

• “standard_*” detection mode: Designed for text detection, supports Simplified Chinese, Traditional Chinese, English, Japanese, French, German, Italian, Spanish, Portuguese, Dutch, Polish, Romanian, Czech, Swedish, Norwegian, Danish, Finnish, Hungarian, Turkish, Vietnamese, Indonesian, Malay, Azerbaijani, Afrikaans, Bosnian, Croatian, Welsh, Estonian, Irish, Icelandic, Kurdish, Lithuanian, Latvian, Maltese, Maori, Occitan, Slovak, Slovenian, Albanian, Swahili, Tagalog, Uzbek, Latin, Serbian (Latin), Catalan, Basque, Galician, Luxembourgish, Romansh, Quechua.

  tip

  Language support indicates detectable text, but does not guarantee consistent results for all fonts, artistic text, or low-contrast text.

• precise detection mode: Based on large models for target recognition, supports identification and inpainting of text in various common languages.

Supported regions #

• Beijing: cn-beijing

To use the operator in the Shanghai or Shenzhen region, submit a ticket to contact technical support for evaluation and assistance.

Operator performance #

Input requirements #

Output requirements #

• 计费标准

  细分项	计费标准说明
  计费项	基于输出视频的时长进行用量统计，计费时，会基于输出视频的帧率、分辨率，对输出视频时长进行系数折算，折算后的计费时长作为计费用量。 任务处理完成后，您可通过 poll 接口返回的 output_fps、resolution 参数查看当前任务实际输出视频的帧率、分辨率取值。
  计费类型	按量计费，单位：元/分钟，按实际的计费用量每小时出账。计费精度：分钟计费，精确到毫秒级
  单价	单价与您选择使用视频修复时的修复版本、目标检测和修复模式有关。提交算子任务时，支持通过 model_level、detection_mode 请求参数设置任务的修复版本、目标检测和修复模式

  warning

  • 支付的总费用：除单价外，服务计费与实际修复后视频时长和视频帧率、分辨率关联。
  • 视频修复不改变视频帧率，修复后的视频时长可能会小于输入的视频时长。

• 计费详情
  计费公式：总费用 = 单价 * 计费用量 = 单价 * (视频时长 * 时长-输出帧率折算系数 * 时长-输出分辨率折算系数)

  细分场景	单价	计费时长-输出帧率折算系数	计费时长-输出分辨率折算系数
  极速版修复，即 model_level : lite	0.5 元/分钟	实际输出帧率： (0 , 24fps]：系数为 0.8 (24fps , 30fps]：系数为 1 大于 30fps：将按照实际输出视频的帧率与基础帧率（30fps）的倍率（向上取整）作为系数	实际输出总像素数： (0 , 1920×1080]：系数为 1 (1920×1080 , 3840×2160]：系数为 2
  专业版 - standard 级目标检测与修复，即 model_level : pro detection_mode : standard_*	1.5 元/分钟		实际输出总像素数 ： (0 , 1280×720]：系数为 0.6 (1280×720 ，1920×1080]：系数为 1 (1920×1080 , 3840×2160]：系数为 2
  专业版 - precise 级目标检测与修复，即 model_level : pro detection_mode : precise	3 元/分钟

• 计费示例
  使用专业版 - precise 级目标检测与修复模式进行视频修复，修复后实际输出视频的帧率为 50fps 、分辨率为 1080p（1920×1080）视频时长为 10 分钟时：

  • 实际输出视频帧率与基础帧率的倍率为1.6，向上取整后为 2.0。
  • 实际费用为：
    总费用 = 3 元/分钟 * （10分钟 * 2 * 1）= 60 元 。

API invocation instructions #

• The API of this operator is an asynchronous task interface. You need to submit a task via the Submit interface to obtain the task's task_id, then use the Poll interface to check the status of the corresponding task (task_id) and retrieve the operator's processing result.
• Caution: After a task is submitted, the task_id is valid for 3 days. After 3 days, you will not be able to retrieve the operator's processing result using this task_id.

Submit #

Interface description #

Submit an asynchronous Video Inpaint task. The operator reads a video, determines the inpainting range based on detection mode or fixed region, outputs the inpainted video, and retains the original audio.

Request parameters #

Response parameters #

Example #

Request example

export OUTPUT_TOS_PATH="tos://path/to/output/"

curl --location "https://operator.las.cn-beijing.volces.com/api/v1/submit" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer $LAS_API_KEY" \
--data '{
  "operator_id": "las_video_inpaint_pro",
  "operator_version": "v1",
  "data": {
    "video_url": "https://${LAS_TOS_BUCKET}.${TOS_ENDPOINT}/operator_cards_serving/public/${LAS_ENVIRONMENT}/las_video_inpaint_pro/v1/input.mp4",
    "output_tos_path": "$OUTPUT_TOS_PATH",
    "model_level": "pro",
    "detection_mode": "precise",
    "targets": ["watermark", "subtitle"],
    "maximum_output_fps": 30,
    "maximum_resolution": "1080p",
    "output_video_mode": "balanced"
  }
}'

Response example

{
  "metadata": {
    "task_id": "task-20260812154600-abc123",
    "task_status": "PENDING",
    "business_code": "0",
    "error_msg": ""
  }
}

Error code #

Poll #

Interface description #

Query the status of the Video Inpaint task using the task_id returned by Submit. After the task is completed, data returns the result address and information about the regions requested to be enabled and actually detected.

Request parameters #

Response parameters #

Example #

Request example

curl --location "https://operator.las.cn-beijing.volces.com/api/v1/poll" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer $LAS_API_KEY" \
--data '{
  "operator_id": "las_video_inpaint_pro",
  "operator_version": "v1",
  "task_id": "task-20260812154600-abc123"
}'

Response example

{
  "metadata": {
    "task_id": "task-20260812154600-abc123",
    "task_status": "COMPLETED",
    "business_code": "0",
    "error_msg": ""
  },
  "data": {
    "inpainted_video_path": "tos://bucket/output/video_inpaint/xxx/output.mp4",
    "output_video_url": "https://example.com/output.mp4",
    "subtitle_bbox": "120,800,1800,960",
    "resolution": "1920x1080",
    "output_fps": 29.97
  }
}

Error code #

• HTTP 200 only indicates that the Poll request was successful; it does not mean the task was successful. If the task fails, check metadata.task_status, metadata.business_code, and metadata.error_msg.