SparkVox Draft Pipeline: Judgment, Not Summaries

Most AI repurposing tools stop at summarization: turn a transcript into posts until you hit a count. SparkVox treats the hard part as editorial judgment - which moments are worth saying publicly, which hook wins, and whether a draft sounds like anyone in your category. That is the pipeline change worth understanding if you care about voice, not volume.

Before a post exists: curation

Sparky scores candidate moments (relevance 70+), then keeps a subset for your recording length - roughly eight posts per 30 minutes, cap 15. The ready screen shows moments found vs posts curated so you see the filter in action. See why SparkVox curates fewer posts per project.

Curation also learns from behavior: topics and post styles you often edit heavily or discard can be deprioritized on the next project. Discarding is a stronger signal than a light edit - it means the material should not have been surfaced. How discard signals feed the next project.

While drafting: two candidates, one card

For each curated moment, Sparky generates two LinkedIn drafts in parallel and keeps the one with the stronger hook and readiness score. You still review one card per moment - the extra work happens before your queue, not after. Why two drafts per moment.

Awareness and Authority posts can also pass a distinctiveness check so generic category filler is less likely to reach you. Conversion posts lean on proof and buyer recognition; the gate focuses on the stages where sameness hurts most. See distinctiveness check and generic AI.

Context that steers judgment

Every generation still uses Brief Sparky, Project Presets, your compounding voice profile, and top-performing publishes when Post Analytics has data. The optional positioning claim on Brief Sparky adds a one-sentence thesis: Sparky uses it when picking moments and framing transcript lines as evidence, not recap.

After you decide: voice compounds

Approvals and edits update your voice profile. Voice Health can cluster repeated edits into automatic rules. The Voice Profile drawer Draft Feedback tab and Analytics theme map show what Sparky learned from your decisions - publish vs edit vs discard by topic and funnel column. Voice Confidence explained covers the score; how voice learning works covers the mechanics.

Pipeline reference: how project processing works.