@scikkk

sam segments images using meta's segment anything model to extract foreground subjects as transparent pngs, with support for hint points, multi-object extraction, and three model sizes trading speed for quality.

lyric-flip extracts rhythmic and rhyme constraints from a reference song, then generates new lyrics adhering to that skeleton before composing audio via senseaudio api. works for parodies and structured rewrites in chinese and english.

使用 SAM（Segment Anything Model）去除图像背景，将前景主体提取为透明 PNG。适用于去除背景、抠图、提取前景主体或图像分割等需求。

Fix Chinese polyphone (多音字) mispronunciation in TTS by auto-detecting ambiguous characters and applying pinyin annotations. Use when users complain about wro...