guyaga

10d10s Bonus Ultimate Subs Maker — AI skill for Claude Code

AI community

How to install 10d10s Bonus Ultimate Subs Maker

This entry records only its repository, not the path inside it, so there is no exact command to give. Open guyaga/10d10s-bonus-ultimate-subs-maker and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What 10d10s Bonus Ultimate Subs Maker does

Claude Code skill - Hebrew subtitles that are in sync and spelled right: Whisper (ivrit.ai model) times every word, Gemini watches the video and fixes names, UI labels and English terms, Claude reviews, then SRT + ASS + a burned-in copy with optional audio clean-up. Course bonus for 10 Days 10 Skills.

Alternatives in AI

Azure OpenAI .net — GPT-4, embeddings, DALL-E, and Whisper client 1.8k ★
Dotai — The ultimate AI development stack, including Claude Code, Task Master, and Curso 1.1k ★
Update Branch Name — by giselles-ai - Updates branch names with proper prefixes and formats, enforcing naming conventions, supporti 502 ★

README

Ultimate Subs Maker: a Claude Code skill

**Hebrew subtitles that are in sync and spelled right. Whisper sets the timing, Gemini watches the video and fixes the words, Claude checks what's left.**

![cover](assets/cover.jpg)

Whisper times every word but mishears names and tech terms. Gemini reads the screen and spells them right, but its own timestamps drift (in our test, 80 seconds off by minute 9 of a 13-minute video). This skill uses each one for what it does best:

Whisper (the ivrit.ai Hebrew model, on your GPU) transcribes with a timestamp for every word
Gemini watches the video with the transcript and fixes only the wording: product names, UI labels, English terms, punctuation. It never touches the timing
The corrected words snap back onto Whisper's timings, so nothing drifts
Claude reviews a change report and the words Gemini was unsure of, before anything is burned
Cues of up to 2 lines × 42 characters, with RTL marks so mixed Hebrew/English lines never flip
Output: SRT for YouTube/Vimeo, a styled ASS, and a new copy of the video with the subtitles burned in. Optional audio clean-up to -16 LUFS. The original video is never touched

כתוביות בעברית שמסונכרנות ומאויתות נכון. Whisper עם המודל העברי של ivrit.ai נותן תזמון לכל מילה, Gemini צופה בסרטון ומתקן את הניסוח (שמות, כפתורים, מונחים באנגלית) בלי לגעת בתזמון, ו-Claude עובר על השינויים לפני הצריבה. בסוף מקבלים SRT, קובץ ASS מעוצב ועותק חדש של הסרטון עם הכתוביות צרובות, ואם רוצים גם סאונד משופר.

Install

Install the ultimate-subs-maker skill from https://github.com/guyaga/10d10s-bonus-ultimate-subs-maker by cloning it to
~/.claude/skills/ultimate-subs-maker, then run python ~/.claude/skills/ultimate-subs-maker/scripts/setup.py. It installs
the packages, ffmpeg, the Hebrew Whisper model (about 3 GB) and the GPU libraries it needs. Fix anything it marks [!!]:
if GEMINI_API_KEY is missing, ask me for it and save it as GEMINI_API_KEY=... in ~/.ultim