Benchmark speech-to-text models (Whisper, faster-whisper) for this GPU beside the resident local LLM, all on VRAM, and print a scored table to pick the best one: transcript error against your own reference words, live-caption speed, and how much VRAM each takes and pushes the LLM out of.
Installs into .claude/skills of the current project.
Are you the author of Opt Local Stt Vram?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/lari-uqac-opt-local-stt-vram)
---
name: opt-local-stt-vram
description: 'Benchmark speech-to-text models (Whisper, faster-whisper) for this GPU beside the resident local LLM, all on VRAM, and print a scored table to pick the best one: transcript error against your own reference words, live-caption speed, and how much VRAM each takes and pushes the LLM out of.'
---
Pointer mirror generated by `install.ps1`. The authoritative skill, with every
instruction, script and reference it ships, lives at
`.claude/skills/opt-local-stt-vram/SKILL.md` in this repository.
MANDATORY FIRST STEP: read `.claude/skills/opt-local-stt-vram/SKILL.md` in full and follow
it exactly. Do not act from this pointer alone - it carries the trigger description
and nothing else.
