Validating: AI dictation prompt packs for domain-specific voice workflows
Background: I use AI dictation (Superwhisper + Wispr Flow) for most of my writing. The problem: the same dictation settings that work great for quick emails produce terrible output for technical docs, legal drafts, or academic writing.
I spent a few weeks building custom context prompts for each use case. Each prompt primes the AI with vocabulary style, punctuation rules, output format, and domain conventions before I start speaking.
The result: substantially better first-draft quality with zero post-editing for casual writing, and about 30% less cleanup on technical content.
Use cases I've built prompts for so far:
- Legal drafting (formal, citation-ready)
- Software engineering notes (code-friendly formatting)
- GTD/inbox processing (action verb sentences, no filler)
- Email (adaptive formality based on recipient type)
- Academic writing (citation format, hedged language)
- Content creation (hook-first structure)
I'm testing whether to productize this as a $19 one-time Gumroad pack — plain text files, drop-in compatible with Wispr Flow, SuperWhisper, and any app that accepts custom prompts. Zero subscription, you own it.
Is this a real friction point for others, or just me? Would a pre-built set of these prompts be worth paying for?
This is a real friction point, but I’d be careful positioning it as just a “prompt pack.” The stronger idea is workflow-specific voice intelligence. People do not only want better dictation; they want spoken input to land in the right professional format without constantly reworking tone, structure, citations, or cleanup.
The best wedge is probably one high-value use case first, like legal drafting, technical notes, or academic writing, instead of selling a broad bundle too early. That makes the pain clearer and lets buyers feel, “this was built for my exact workflow.”
If this turns into more than a Gumroad pack and becomes a serious voice-workflow product, Lyriso.com could fit the softer productivity and writing-support direction better than a generic prompt-pack name.