AI Skill MartBrowse skillsPromptsProgramsLearnWhat is a Skill?
 
← Back to list

Transcript Fixer

Cleans up speech-to-text transcripts through a two-stage pipeline — instant dictionary rules plus Claude's own AI correction — while building a personal correction database that gets smarter with every fix.

UtilitiesIntermediate★ 1,339⑂ 214AI score 9/10Last updated: Aug 16, 2026

What it does

  • Stage 1 (dictionary): applies stored SQLite correction rules instantly and for free. Safe mode is the default, so risky rules are written to a *_needs_review.md sidecar instead of being applied silently.
  • Stage 2 (native AI correction, default, no API key): Claude reads the transcript itself and fixes proper nouns, domain jargon and context-dependent homophones — the pass that does most of the real work on clean ASR.
  • Learning loop: confirmed fixes are --added to a domain dictionary, recurring people go into a people roster, and ambiguous "traps" go into per-domain context files that prime the AI pass.
  • Review queue: uncertain items persist in the database with evidence and action packs, guarded by a fail-closed anchor check that refuses to auto-edit when the file has drifted.
  • Safety tooling: --dry-run previews, always-on change reports, --audit for risky rules, --report-false-positive, --probe corpus evidence, and a stable --json status contract for caller pipelines.

Who it's for

  • Anyone who regularly edits meeting notes, interviews or lecture recordings
  • Teams post-processing Whisper / Otter / conferencing auto-captions
  • Users tired of correcting the same product or person name every single time
  • Automation builders wiring transcript cleanup into an ingest pipeline

Examples

  1. Run --init once, then --input meeting.md --stage 1, and let Claude read the output to fix the remaining recognition errors before you file it.
  2. For a curated project vocabulary, run --domain myproject --apply-domain so all hand-confirmed rules land in a single pass instead of three reruns.
  3. Save a confirmed pattern with --add "cloucode" "Claude Code" --domain tech so the next transcript fixes it automatically.
  4. Batch-process a folder with a shell loop for Stage 1, then hand each file to the native AI pass.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/daymade/claude-code-skills/HEAD/daymade-audio/transcript-fixer/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the daymade-audio/transcript-fixer folder from the GitHub repo daymade/claude-code-skills into my ~/.claude/skills/transcript-fixer/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/daymade/claude-code-skills.git && mkdir -p ~/.claude/skills && cp -r claude-code-skills/daymade-audio/transcript-fixer ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Install uv first — the scripts use PEP 723 inline metadata and let uv auto-install dependencies: curl -LsSf https://astral.sh/uv/install.sh | sh
  2. Clone the repository: git clone https://github.com/daymade/claude-code-skills.git
  3. Copy the skill into place: mkdir -p ~/.claude/skills && cp -r claude-code-skills/daymade-audio/transcript-fixer ~/.claude/skills/
  4. Restart Claude Code and confirm transcript-fixer appears in your skill list.
  5. Initialize the database once: from the skill directory run uv run scripts/fix_transcription.py --init
  6. Use it conversationally, e.g. "clean up the ASR errors in these meeting notes". Applied: 0 from Stage 1 is expected under safe mode — the job is only done after the native AI pass and --add of confirmed fixes.
  7. Optional: only for agent-less CI batches, export GLM_API_KEY=... and use --stage 3; inside Claude Code you should not run Stage 3.
View source on GitHub ↗License: MIT