wedding-india
Indian wedding film editor for Claude: sorts multi-day footage by event (Haldi, Mehndi, Sangeet, baraat, pheras or Nikah, vidaai, reception) and cuts a 3-7 min cinematic highlight, per-event 9:16 reels, a fast teaser and a long traditional film, with respectful ritual pacing and
Install
npx skills add https://github.com/Viprasol-Tech/claude-video-editing-skills/tree/main/plugins/video-studio/skills/wedding-india
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install viprasol-tech-claude-video-editing-skills@llmmart
git clone https://github.com/Viprasol-Tech/claude-video-editing-skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole viprasol-tech/claude-video-editing-skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Wedding India
Several days, five or six functions, hundreds of relatives: one film per event, one highlight that carries the whole celebration, and the rituals cut with the respect the family expects.
- Best for: Hindu, Sikh, Muslim, Jain and mixed Indian weddings; destination weddings; engagement (roka, sagai)
- Inputs: footage per function (ideally one folder per event), photos, the couple's song choices, mantra / vow audio
- Default skin:
soft-gradient· alternativescinematic-phone,frame-studio,minimal-mono - Formats: 16:9 highlight and full film, 9:16 per-event reels and teaser
- Durations: teaser 30-60 s (within days), per-event reel 30-90 s, highlight 3-7 min, traditional film 1-3 h
- Genre preset:
--genre wedding-india(python CORE/scripts/editor.py genre wedding-india)
Before you start
- Needs video-studio-core in the same skills folder.
CORE= that folder,SKILLS= the skills folder. - First run on a machine:
python CORE/scripts/check_env.py. - Ask the family first and write it down: couple's names in English and in the script they want (Hindi, Gujarati, Tamil, Punjabi, Urdu...), event names and dates, the list of elders who must appear (grandparents, both sets of parents, mama/mami, bua...), the order they want, songs (licensed), and anything that must not be shown.
- Consent and privacy: many guests, women in purdah, children. Deliver privately; public posting only with the couple's approval; drop anyone who asked not to be filmed.
- Sort footage into one folder per event before editing (
haldi/,mehndi/,sangeet/,baraat/,pheras/,vidaai/,reception/) and give each event its own work folder. Never mix events in one assembly.
How a pro cuts this
Highlight structure (% of length)
| beat | % | what goes here |
|---|---|---|
| haldi | 12 | turmeric close-ups, laughter, water / flowers thrown |
| mehndi | 12 | henna detail macros, the groom's name hidden in the design, dholki |
| sangeet | 18 | family performances, the couple's dance, peak energy on the drop |
| baraat / arrival | 12 | horse or car, dhol, dancing relatives, the varmala |
| pheras / nikah | 26 | the ritual heart: agni, mantras, the seven rounds, sindoor, mangalsutra; or qubool hai and the nikahnama |
| vidaai / reception | 20 | the emotional goodbye, then the stage, cake, the last dance |
Per-event reel: open on the strongest image of that event, 3-4 beats of build, end on a family moment. Same skin and grade for every event so the set looks like one wedding.
Ritual pacing. Ceremonies are repetitive but sacred. Never cut the seven pheras into seven 1-second flashes: show the first round in full (or most of it), dissolve through the middle rounds, hold the last round and the reaction. Let a mantra or the qubool hai play with the music dipped right down, or none. Hold the sindoor and mangalsutra moments 3-5 s. Cut the celebrations on the beat (ASL 1.5-3.5 s), the rituals on the phrase (4-8 s holds).
Cut rules. No jump cuts inside a ritual; cover with a different angle or a detail (hands, fire, flowers). Never place a laughing face over a crying vidaai moment (Kuleshov). Never reorder a ritual: the family knows the order. Every elder on the list appears at least once, ideally in a clean close-up.
Transition grammar. Cuts on the beat for haldi, mehndi, sangeet, baraat; slow dissolves (0.8-1.2 s) inside pheras and vidaai; dip to black between events in the long film. Whip only in sangeet / baraat.
Music and sound. A different licensed song per event is normal; the highlight can carry two songs joined on a downbeat. Song leads (bed ~-6 dB) and dips fully for mantras, vows and the vidaai. Keep natural dhol and crowd sound.
Colour. indian-wedding (warm gold, rich reds, skin kept natural) at 0.6-0.75; never push reds so far the bridal
lehenga clips. Haldi yellow should look yellow, not orange skin.
Titles and captions. Event names as titles in the family's script (हल्दी, मेहंदी, संगीत, फेरे; or Tamil, Gujarati...). No creator captions in the highlight. Mantra translation subtitles only if the family asks.
What the client judges. Every elder present, ritual respect, colour of the outfits, the sangeet energy, the teaser arriving within days.
Common mistakes. One assembly for all events; pheras rushed; a missing grandparent; orange skin from haldi; music over the mantras; misspelled names in the regional script; boxes instead of letters (font without that script).
Make the edit
One work folder per event. Montage orders clips by file modification time and bakes the song into the assembled audio, so dry-run first and fix the order in the spec.
# per event, e.g. sangeet (beat-cut montage, 60 s reel in 9:16)
python CORE/scripts/timeline.py auto sangeet/ --work ev-sangeet --mode montage --music sangeet_song.wav --duration 60 --format 9:16 --dry-run
# edit ev-sangeet/timeline_spec.json: order, drop shots, transitions
python CORE/scripts/run.py - --timeline ev-sangeet/timeline_spec.json --style SKILLS/soft-gradient --genre wedding-india --format 9:16 --no-cut --model base --work ev-sangeet --stop-after plan
python CORE/scripts/render.py ev-sangeet/plan.json --frames 1,10,25,45,58
python CORE/scripts/render.py ev-sangeet/plan.json
python CORE/scripts/finish.py ev-sangeet/plan.json --platform reels
Pheras / Nikah (ritual sound kept). Use a hand spec (or timeline.py auto pheras/ --mode story --model base --dry-run)
with the ritual clips "audio": "keep", dissolves between rounds, and no song baked in. Transcribe in the spoken language:
python CORE/scripts/timeline.py auto pheras/ --work ev-pheras --mode story --format 16:9 --model base --dry-run
python CORE/scripts/run.py - --timeline ev-pheras/timeline_spec.json --style SKILLS/soft-gradient --genre wedding-india --format 16:9 --no-cut --model base --music pheras_bed.wav --work ev-pheras --stop-after plan
# run.py does not pass --language to the timeline transcription; redo it in the spoken language, then re-plan:
python CORE/scripts/transcribe.py --work ev-pheras --model small --language hi
python CORE/scripts/autocut.py --work ev-pheras --keep-all
python CORE/scripts/make_plan.py --work ev-pheras --style SKILLS/soft-gradient --genre wedding-india --format 16:9
Re-planning drops the --music that run.py added, so put the music block back in the plan by hand.
Highlight: build the spec from the best shots of every event in ritual order (haldi to reception), then
run.py - --timeline highlight_spec.json ... --format 16:9 --no-cut --work hl. Traditional film: the same per event
with --mode story, long holds and dip-to-black between events.
Refine the plan
- Remove creator captions (delete the
captionselement the skin adds) and clearwordsin montage reels. - Montage audio is music:
"audio": {"platform": "youtube", "voice": false}. - Ritual sections:
"music": {"src": "pheras_bed.wav", "under": 22, "duck": "speech-aware", "duckDepth": 20}keeps a soft bed that drops away under the mantras. Limit: plan music runs under the whole video (startis an offset into the song, not a timeline position) and there is no per-section mute. When a mantra must stand completely alone, leave music out of the pheras film and keep it for the celebration reels. - Colour:
"color": {"look": "indian-wedding", "intensity": 0.7}; mixed cameras"match": "hero.png"."finish": {"grain": 3, "vignette": 0.25}. - Indic-script titles: load a font that has the script, in the plan theme:
"theme": {"fontsUrl": "https://fonts.googleapis.com/css2?family=Noto+Sans+Devanagari:wght@600&display=swap", "vars": {"--font-display": "'Noto Sans Devanagari', sans-serif"}}(Tamil:Noto Sans Tamil, Gujarati:Noto Sans Gujarati, Gurmukhi:Noto Sans Gurmukhi, Urdu:Noto Nastaliq Urdu). Keep the skin'sfontsUrltoo if you need its Latin face (one URL can load severalfamily=entries). Title:{"type": "titlecard", "kicker": "16.02.2026", "title": "संगीत", "subtitle": "Riya & Arjun", "start": 0, "end": 3}. - Event lower third for elders (optional):
{"type": "lowerthird", "name": "Dadi ji", "role": "Grandmother", ...}, only with names the family approved. - Freeze or slow the varmala / sindoor moment in the spec:
"speedRamp": [{"at": 12.0, "rate": 0.5, "ramp": 0.4}]. python CORE/scripts/editor.py review ev-sangeet/plan.json --genre wedding-india.- Footage clean-up (video, before the spec): handheld baraat shots, dark sangeet halls and LED-lit stages:
python CORE/scripts/prep.py baraat03.mp4 --work fix/baraat03 --stabilize --denoise light, then usefix/baraat03/source.mp4in the spec.--stabilize= vidstab (crops ~6% at the edges),--denoise light|strong= video noise,--deflicker= mains-light flicker. Only on clips that need it.--deflickerunder tube/LED lights that pulse on camera.
Deliver
python CORE/scripts/deliver.py hl/plan.json --to youtube,prores-422hq --out-dir hl/deliver --burn no
python CORE/scripts/deliver.py ev-sangeet/plan.json --to reels,whatsapp-status --out-dir ev-sangeet/deliver --burn no
python CORE/scripts/qc.py hl/deliver/final_youtube.mp4 --plan hl/plan.json --platform youtube --strict
python CORE/scripts/qc.py ev-sangeet/deliver/final_reels.mp4 --plan ev-sangeet/plan.json --platform reels --strict
whatsapp-status makes a <=16 MB, 60 s version families forward. Sidecar captions.srt files are UTF-8 and keep the
regional script; check they open correctly. Name files riya-arjun_sangeet_v001_9x16.mp4.
Check before delivering
- Contact sheet: the regional script renders with correct conjuncts and matras (no boxes, no broken ि placement).
If it is wrong, the font lacks the script: fix
fontsUrl. - Every elder on the family's list ticked off; names in both scripts spelled as the family wrote them.
- Pheras / Nikah: rituals in the correct order, mantras audible, no music over them, no rushed rounds.
- Skin natural under haldi yellow and stage lights; reds not clipping.
qc.pypasses loudness and flash checks. - Tell the family what each file is, which events are done, and how many correction rounds remain (often 2).
Example requests
- "Make a 60-second Sangeet reel and a Haldi reel from these folders, Bollywood song for each."
- "Cut a 5-minute highlight of the whole wedding with the event names in Hindi."
- "Edit the pheras properly, keep the mantras, don't rush the rounds."
Files (claude-video-editing-skills)
-
SKILL.md 11.1 KB
--- name: wedding-india description: "Indian wedding film editor for Claude: sorts multi-day footage by event (Haldi, Mehndi, Sangeet, baraat, pheras or Nikah, vidaai, reception) and cuts a 3-7 min cinematic highlight, per-event 9:16 reels, a fast teaser and a long traditional film, with respectful ritual pacing and titles in Devanagari, Tamil, Gujarati or other Indic scripts. Use when the user says 'edit my shaadi video', 'Indian wedding highlight', 'make a Sangeet reel', 'Haldi reel', 'cut the pheras', 'Nikah film', 'traditional wedding film', 'wedding teaser in Hindi', or hands over footage from several wedding functions. Runs on video-studio-core (timeline, beats, colour, deliver, QC). For a single-day Western wedding use wedding-highlight." license: MIT --- # Wedding India Several days, five or six functions, hundreds of relatives: one film per event, one highlight that carries the whole celebration, and the rituals cut with the respect the family expects. - **Best for:** Hindu, Sikh, Muslim, Jain and mixed Indian weddings; destination weddings; engagement (roka, sagai) - **Inputs:** footage per function (ideally one folder per event), photos, the couple's song choices, mantra / vow audio - **Default skin:** `soft-gradient` · alternatives `cinematic-phone`, `frame-studio`, `minimal-mono` - **Formats:** 16:9 highlight and full film, 9:16 per-event reels and teaser - **Durations:** teaser 30-60 s (within days), per-event reel 30-90 s, highlight 3-7 min, traditional film 1-3 h - **Genre preset:** `--genre wedding-india` (`python CORE/scripts/editor.py genre wedding-india`) ## Before you start - Needs **video-studio-core** in the same skills folder. `CORE` = that folder, `SKILLS` = the skills folder. - First run on a machine: `python CORE/scripts/check_env.py`. - Ask the family first and write it down: couple's names in English **and** in the script they want (Hindi, Gujarati, Tamil, Punjabi, Urdu...), event names and dates, the list of elders who must appear (grandparents, both sets of parents, mama/mami, bua...), the order they want, songs (licensed), and anything that must not be shown. - **Consent and privacy:** many guests, women in purdah, children. Deliver privately; public posting only with the couple's approval; drop anyone who asked not to be filmed. - Sort footage into one folder per event **before** editing (`haldi/`, `mehndi/`, `sangeet/`, `baraat/`, `pheras/`, `vidaai/`, `reception/`) and give each event its own work folder. Never mix events in one assembly. ## How a pro cuts this **Highlight structure (% of length)** | beat | % | what goes here | | --- | --- | --- | | haldi | 12 | turmeric close-ups, laughter, water / flowers thrown | | mehndi | 12 | henna detail macros, the groom's name hidden in the design, dholki | | sangeet | 18 | family performances, the couple's dance, peak energy on the drop | | baraat / arrival | 12 | horse or car, dhol, dancing relatives, the varmala | | pheras / nikah | 26 | the ritual heart: agni, mantras, the seven rounds, sindoor, mangalsutra; or qubool hai and the nikahnama | | vidaai / reception | 20 | the emotional goodbye, then the stage, cake, the last dance | Per-event reel: open on the strongest image of that event, 3-4 beats of build, end on a family moment. Same skin and grade for every event so the set looks like one wedding. **Ritual pacing.** Ceremonies are repetitive but sacred. Never cut the seven pheras into seven 1-second flashes: show the first round in full (or most of it), dissolve through the middle rounds, hold the last round and the reaction. Let a mantra or the qubool hai play with the music dipped right down, or none. Hold the sindoor and mangalsutra moments 3-5 s. Cut the celebrations on the beat (ASL 1.5-3.5 s), the rituals on the phrase (4-8 s holds). **Cut rules.** No jump cuts inside a ritual; cover with a different angle or a detail (hands, fire, flowers). Never place a laughing face over a crying vidaai moment (Kuleshov). Never reorder a ritual: the family knows the order. Every elder on the list appears at least once, ideally in a clean close-up. **Transition grammar.** Cuts on the beat for haldi, mehndi, sangeet, baraat; slow dissolves (0.8-1.2 s) inside pheras and vidaai; dip to black between events in the long film. Whip only in sangeet / baraat. **Music and sound.** A different licensed song per event is normal; the highlight can carry two songs joined on a downbeat. Song leads (bed ~-6 dB) and dips fully for mantras, vows and the vidaai. Keep natural dhol and crowd sound. **Colour.** `indian-wedding` (warm gold, rich reds, skin kept natural) at 0.6-0.75; never push reds so far the bridal lehenga clips. Haldi yellow should look yellow, not orange skin. **Titles and captions.** Event names as titles in the family's script (हल्दी, मेहंदी, संगीत, फेरे; or Tamil, Gujarati...). No creator captions in the highlight. Mantra translation subtitles only if the family asks. **What the client judges.** Every elder present, ritual respect, colour of the outfits, the sangeet energy, the teaser arriving within days. **Common mistakes.** One assembly for all events; pheras rushed; a missing grandparent; orange skin from haldi; music over the mantras; misspelled names in the regional script; boxes instead of letters (font without that script). ## Make the edit One work folder per event. Montage orders clips by file modification time and bakes the song into the assembled audio, so dry-run first and fix the order in the spec. ```bash # per event, e.g. sangeet (beat-cut montage, 60 s reel in 9:16) python CORE/scripts/timeline.py auto sangeet/ --work ev-sangeet --mode montage --music sangeet_song.wav --duration 60 --format 9:16 --dry-run # edit ev-sangeet/timeline_spec.json: order, drop shots, transitions python CORE/scripts/run.py - --timeline ev-sangeet/timeline_spec.json --style SKILLS/soft-gradient --genre wedding-india --format 9:16 --no-cut --model base --work ev-sangeet --stop-after plan python CORE/scripts/render.py ev-sangeet/plan.json --frames 1,10,25,45,58 python CORE/scripts/render.py ev-sangeet/plan.json python CORE/scripts/finish.py ev-sangeet/plan.json --platform reels ``` **Pheras / Nikah (ritual sound kept).** Use a hand spec (or `timeline.py auto pheras/ --mode story --model base --dry-run`) with the ritual clips `"audio": "keep"`, dissolves between rounds, and no song baked in. Transcribe in the spoken language: ```bash python CORE/scripts/timeline.py auto pheras/ --work ev-pheras --mode story --format 16:9 --model base --dry-run python CORE/scripts/run.py - --timeline ev-pheras/timeline_spec.json --style SKILLS/soft-gradient --genre wedding-india --format 16:9 --no-cut --model base --music pheras_bed.wav --work ev-pheras --stop-after plan # run.py does not pass --language to the timeline transcription; redo it in the spoken language, then re-plan: python CORE/scripts/transcribe.py --work ev-pheras --model small --language hi python CORE/scripts/autocut.py --work ev-pheras --keep-all python CORE/scripts/make_plan.py --work ev-pheras --style SKILLS/soft-gradient --genre wedding-india --format 16:9 ``` Re-planning drops the `--music` that run.py added, so put the `music` block back in the plan by hand. **Highlight:** build the spec from the best shots of every event in ritual order (haldi to reception), then `run.py - --timeline highlight_spec.json ... --format 16:9 --no-cut --work hl`. **Traditional film:** the same per event with `--mode story`, long holds and dip-to-black between events. ## Refine the plan - **Remove creator captions** (delete the `captions` element the skin adds) and clear `words` in montage reels. - **Montage audio is music:** `"audio": {"platform": "youtube", "voice": false}`. - **Ritual sections:** `"music": {"src": "pheras_bed.wav", "under": 22, "duck": "speech-aware", "duckDepth": 20}` keeps a soft bed that drops away under the mantras. Limit: plan music runs under the whole video (`start` is an offset into the song, not a timeline position) and there is no per-section mute. When a mantra must stand completely alone, leave music out of the pheras film and keep it for the celebration reels. - **Colour:** `"color": {"look": "indian-wedding", "intensity": 0.7}`; mixed cameras `"match": "hero.png"`. `"finish": {"grain": 3, "vignette": 0.25}`. - **Indic-script titles:** load a font that has the script, in the plan theme: `"theme": {"fontsUrl": "https://fonts.googleapis.com/css2?family=Noto+Sans+Devanagari:wght@600&display=swap", "vars": {"--font-display": "'Noto Sans Devanagari', sans-serif"}}` (Tamil: `Noto Sans Tamil`, Gujarati: `Noto Sans Gujarati`, Gurmukhi: `Noto Sans Gurmukhi`, Urdu: `Noto Nastaliq Urdu`). Keep the skin's `fontsUrl` too if you need its Latin face (one URL can load several `family=` entries). Title: `{"type": "titlecard", "kicker": "16.02.2026", "title": "संगीत", "subtitle": "Riya & Arjun", "start": 0, "end": 3}`. - **Event lower third for elders (optional):** `{"type": "lowerthird", "name": "Dadi ji", "role": "Grandmother", ...}`, only with names the family approved. - Freeze or slow the varmala / sindoor moment in the spec: `"speedRamp": [{"at": 12.0, "rate": 0.5, "ramp": 0.4}]`. - `python CORE/scripts/editor.py review ev-sangeet/plan.json --genre wedding-india`. - **Footage clean-up (video, before the spec):** handheld baraat shots, dark sangeet halls and LED-lit stages: `python CORE/scripts/prep.py baraat03.mp4 --work fix/baraat03 --stabilize --denoise light`, then use `fix/baraat03/source.mp4` in the spec. `--stabilize` = vidstab (crops ~6% at the edges), `--denoise light|strong` = video noise, `--deflicker` = mains-light flicker. Only on clips that need it. `--deflicker` under tube/LED lights that pulse on camera. ## Deliver ```bash python CORE/scripts/deliver.py hl/plan.json --to youtube,prores-422hq --out-dir hl/deliver --burn no python CORE/scripts/deliver.py ev-sangeet/plan.json --to reels,whatsapp-status --out-dir ev-sangeet/deliver --burn no python CORE/scripts/qc.py hl/deliver/final_youtube.mp4 --plan hl/plan.json --platform youtube --strict python CORE/scripts/qc.py ev-sangeet/deliver/final_reels.mp4 --plan ev-sangeet/plan.json --platform reels --strict ``` `whatsapp-status` makes a <=16 MB, 60 s version families forward. Sidecar `captions.srt` files are UTF-8 and keep the regional script; check they open correctly. Name files `riya-arjun_sangeet_v001_9x16.mp4`. ## Check before delivering - Contact sheet: the regional script renders with correct conjuncts and matras (no boxes, no broken ि placement). If it is wrong, the font lacks the script: fix `fontsUrl`. - Every elder on the family's list ticked off; names in both scripts spelled as the family wrote them. - Pheras / Nikah: rituals in the correct order, mantras audible, no music over them, no rushed rounds. - Skin natural under haldi yellow and stage lights; reds not clipping. `qc.py` passes loudness and flash checks. - Tell the family what each file is, which events are done, and how many correction rounds remain (often 2). ## Example requests - "Make a 60-second Sangeet reel and a Haldi reel from these folders, Bollywood song for each." - "Cut a 5-minute highlight of the whole wedding with the event names in Hindi." - "Edit the pheras properly, keep the mantras, don't rush the rounds."
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.