Voice-writing a novel: which sub-12B model best adds only punctuation and quotes?
letsburn00 · reddit · 2026-11-14
A Reddit novelist dictates fiction via speech-to-text and wants a model that only adds quotation marks and punctuation — never changing words. Dictating lets each character keep a distinct voice, but manual punctuation passes cost the fresh-reader feel needed to judge the prose's flow.
- Currently uses Whisper Medium with Handy, works well
- Community recs center on Cadence and some form of Gemma, but recent comparisons are scarce
- Constraint: 12GB card, targeting sub-8-12B models, and the author isn't highly technical
A practical discussion of the "format-only fix" niche in voice-writing workflows.
More from coding & agent
- Base44 adds branching for safely testing risky app changes before merge — tristanbob · 2026-11-14
- book-to-skill hits 37.8k stars: turn any technical book PDF into a coding agent skill — tom_doerr · 2026-11-14
- Fable 5.2 solves over half of expert long-horizon engineering tasks, surprising evaluators — Jsevillamol · 2026-11-14
- Dev picks paper as best BYO-agent MCP experience with tens of concurrent subagents — jakedahn · 2026-11-14
- gpucanvas open-sourced: thousands of live DOM nodes on a GPU-drawn canvas — Vjeux · 2026-11-14
- WebGPU canvas renders 7,000 live DOM frames with 2M+ elements at 120fps — Vjeux · 2026-11-14