550W AI Subtitle Remover

Problem

Baked-in subtitles, short-video share-link watermarks, and text on images are hard to clean with simple editing. Raw API calls also introduce unnecessary friction around coordinates, duration, retries, and duplicate paid submissions. This skill turns those operations into natural-language routing: say remove subtitles, remove watermark, or check progress, while the skill chooses the Action and parameters.

How It Works

• Video subtitle removal: supports local files or directly accessible video URLs and always performs full-frame erasure; if audio should be removed, it sets removeAudio: true. Remote direct links are prechecked with ffprobe, and the workflow polls for up to 10 minutes, returning resultUrl, a failure reason, or taskId.
• Video watermark removal: handles short-video platform share links without downloading the video and calls removeVideoWatermark, surfacing data.video from the response.
• Image watermark removal: processes synchronously by default; multiple images are run serially, and async results are tracked via task.taskId and imageWatermarkTaskDetail.
• Tasks and billing: supports taskDetail, taskList, and queryCredits; failed jobs are not charged, and uncertain network timeouts should trigger task queries instead of blind paid retries.

Boundaries

The skill assumes 550W credentials and API-KEY setup. It prompts before batch operations, does not silently expand inputs, asks only one necessary question when intent is unclear, and avoids guessing paid actions.