Borrowing it
Nothing to install: this file belongs to nguyenph88/Media-Editor-MCP. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/nguyenph88/Media-Editor-MCP/main/.claude/skills/pp-erase-text/SKILL.mdgit clone --depth 1 https://github.com/nguyenph88/Media-Editor-MCPWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nguyenph88/media-editor-mcp/pp-erase-text)<a href="https://agentmods.dev/skills/nguyenph88/media-editor-mcp/pp-erase-text"><img src="https://agentmods.dev/badge/skills/nguyenph88/media-editor-mcp/pp-erase-text/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nguyenph88/media-editor-mcp/pp-erase-text"><img src="https://agentmods.dev/badge/skills/nguyenph88/media-editor-mcp/pp-erase-text.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00099 | $0.01055 |
| Opus 5 | $0.00049 | $0.00528 |
| Sonnet 5 | $0.00020 | $0.00211 |
| Haiku 4.5 | $0.00010 | $0.00105 |
Grade A, and why
pp-erase-text scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Erase burned-in captions from a video
Remove static burned-in text (captions/subtitles) from a rendered video and write a clean copy. Pure OpenCV + ffmpeg — no Premiere, no MCP plugin. Audio is copied untouched.
Wraps the standalone script packages/media-analysis/erase_text.py.
How it works (so you tune the right knob)
A caption is a static overlay that persists across a whole shot while the background moves. The script:
- splits the video into shots (PySceneDetect),
- per shot, marks pixels that are white-ish AND persist across ≥
--persistfraction of the shot's frames → one stable mask, - inpaints every frame of that shot with that single mask.
This catches the full caption (even faint/fading frames), rejects moving bright background (whitewater, sky, buildings) via persistence + a glyph-shape filter, and avoids flicker because the mask is constant per shot.
How to run
From packages/media-analysis/ using its venv python:
./.venv/Scripts/python.exe erase_text.py "<path>" [flags]
Key flags:
--preview SEC— write abefore | mask-overlay | resultPNG for the shot at SEC, no video. Always do this first.--out FILE— default<input>_clean.mp4.--band TOP BOT— vertical search region as height fractions (default 0.55 0.88). Move it if captions sit elsewhere.--persist F— fraction of a shot's frames a pixel must be white to count as text (default 0.5). Lower if a caption shows only briefly within a longer shot.--v-min/--s-max— brightness/saturation thresholds for "white" text (defaults 190 / 60).--dilate/--radius— mask growth + inpaint radius (defaults 5 / 5). Bump if residue/edges remain; lower if smudge over busy backgrounds.--max-h/--max-area— drop persistent blobs too tall/large to be text (defaults 0.06 / 0.006).--method telea|ns— inpaint algorithm (default telea).
Workflow
- Find a captioned moment and run
--preview SEC. Read the PNG: the red mask overlay should hug the text and NOT cover background (water/sky/buildings). The right panel shows the inpaint result.- For confidence, also generate a full-res crop of the text band (original above erased) — thumbnails hide residue.
- Tune until the mask is clean across a few representative shots (busy + smooth backgrounds): adjust
--band,--persist,--v-min/--s-max,--dilate. Cheap — preview is one frame. - Full pass — run without
--preview. It logs per-shot mask coverage (and flags shots where no caption was found → passed through untouched). Then encodes. - Validate before telling the user it's openable: the output is MP4 with
+faststart, so a partial/in-progress file is unplayable (DirectShow "MEDIATYPE_Stream / Unknown" error). Wait for theDONElog, then decode-check withffmpeg -v error -i OUT -f null -and extract a captioned frame to confirm text is gone.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 51 lines · 99 tokens per session scan A 2e55e3ae207b
pp-erase-text is a skill published in the GitHub repository nguyenph88/Media-Editor-MCP (0 stars, last pushed 3mo ago), licensed MIT. It adds 99 tokens to every session and 1,055 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
webgl-holographic-foil
A self-contained WebGL2 hero: thin-film interference over a crushed-foil surface whose palette shifts with the viewing angle; move the cursor to tilt the film.
general-video
Author or edit a custom HyperFrames composition when no specialized workflow fits, or when BRIEF.md sets flow: companion. Use for longer or multi-scene pieces, brand and sizzle reels, montages, static loops, static title cards, footage remixes, and freeform builds. Use motion-graphics instead for a short unnarrated…
html-ppt-hermes-cyber-terminal
OpenDesign + BYOK: choosing and wiring your own model, hands-on — cost, quality, and the routing decision. Built as a decision-grade AI literacy deck for engineers, IT, applied-AI teams.
html-ppt-taste-brutalist
16:9 HTML deck in tactical-telemetry / CRT-terminal taste. Deactivated-CRT charcoal slides, white-phosphor monospace, hazard-red accent, scanline overlay, ASCII syntax, density over decoration. Distilled from Leonxlnx/taste-skill brutalist-skill (Tactical Telemetry mode).
chengfeng-check-updates
An environment manager for a video-editing system. It checks whether its skills and runtime—the software needed to run them—are installed and compatible.
diagnostic-stem-delivery
Audio production with diagnostic analysis, timecode parsing from documents, and verified export workflow.