Story-first drafting: AI Song Writer (SongLab) as constrained co-writer

The piece opens by rejecting the pattern most songwriting tools follow — asking for a genre tag and a vague mood, then returning a full lyric set. Its argument is that the bottleneck is not generation at all. Writers can already produce words; the harder parts are turning a personal moment into a singable idea, choosing a structure early enough that verses do not wander, matching lyric tone to musical direction, and iterating without losing the original spark. LLMs are described as good at expansion and variation and weak at taste, which is why the recommended design is "start from your title, feeling, or chorus, then keep the draft movable" rather than one-click output.

The proposed loop — seed, structure, direction, generate options, human edit, save — is presented as runnable in a single sitting, with empty sections counted as progress and two or three candidates kept instead of a final master. Adjacent helpers (vocal remover, stem splitter, key/BPM finder, lyrics generator) sit next to the lyric draft but are best used late, not first. Its product example, AI Song Writer (SongLab), is cited for a split between free local drafts without sign-in and paid credits when AI is invoked, a design framed as matching how people actually write. The closing line is that the writer's voice stays the product and the tool's job is to keep the page from staying blank.