testimonial-film
Customer testimonial editor for Claude: turns a real customer interview into a 60-150 s case-study film (result hook, problem, solution, numbers, recommendation) plus a 30 s cut and vertical clips, with the best soundbite first and every jump cut covered. Use when the user says '
Install
npx skills add https://github.com/Viprasol-Tech/claude-video-editing-skills/tree/main/plugins/video-studio/skills/testimonial-film
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
Testimonial Film
A real customer, in their own words, saying what changed: credible first, polished second.
- Best for: customer case studies, client reviews, agency results, B2B social proof, course student stories
- Inputs: customer interview (one or two angles), B-roll of the customer using the product, result screenshots, logo
- Default skin:
testimonial-card· alternativessplit-proof,quote-card-reel,frame-studio - Formats: 16:9 master, 9:16 and 4:5 cutdowns
- Durations: film 60-150 s, cutdown 30 s, Shorts 15-45 s
- Genre preset:
--genre testimonial-film
Before you start
- Needs video-studio-core in the same skills folder.
CORE= that folder,SKILLS= the skills folder. - First run:
python CORE/scripts/check_env.py. - Ethics (non-negotiable): the speaker is a real customer, the words are theirs, and their meaning is unchanged.
No actors presented as customers, no invented numbers, no
starsrating unless the user supplies the real review. If the customer was paid, got a discount or free product, add a disclosure card (FTC / ASCI). Get the customer's sign-off on the final cut. - Ask for: name, job title, company (spelled exactly), the result numbers and where they come from, logo permission.
How a pro cuts this
Structure (% of length)
| beat | % | what goes here |
|---|---|---|
| result hook | 10 | the single best soundbite: the outcome in their words |
| problem before | 25 | what life was like, the cost of the problem |
| solution | 30 | how they found and used the product, specific moments |
| result / numbers | 25 | concrete results, shown with counters or screenshots |
| recommendation + CTA | 10 | "I'd tell anyone who..." then logo and CTA |
Pacing. ASL 3-6 s, visual change at least every 5 s. Hold the face on emotional or credibility lines; B-roll the explanations.
Cut rules.
- Paper edit from the transcript; the hook may move up from later in the interview (log it), but every sentence stays
whole.
story.py reorderflags partial splices (HIGH) and back-references (MED): resolve every flag. - Cover every join: B-roll >= 0.5 s each side, second angle, or a >= 1.2x punch-in (
speaker.zooms). - Light filler removal; keep natural hesitations that make them sound real. Keep the breath before the key line.
- Never use a reaction or B-roll that changes the meaning (Kuleshov).
Transition grammar. Cuts and J/L cuts; one dissolve at most into the CTA. No flashy transitions.
Music and sound. Bed ~-24 dB under speech, ducked (duck: "speech-aware"), lifted at the recommendation.
Room tone on every join. SFX low (4/min): a soft tick on a counter.
Colour. Natural and trustworthy: documentary-natural or portra at 0.5. Brief says "punchy": clean-commercial 0.6.
Captions. Two-line sentence-case subtitles low in frame ("preset": "clean"), <= 17 cps; burned captions on verticals.
What the client judges. Credibility, the customer sounding natural, a concrete result.
Common mistakes. Starting with "Hi, my name is"; numbers nobody can verify; over-polished audio; obvious jump cuts; music louder than the customer; a quote stitched from two answers.
Make the edit
python CORE/scripts/run.py customer.mp4 --style SKILLS/testimonial-card --genre testimonial-film --format 16:9 --smart-cuts --retakes --logo logo.png --work tf --stop-after plan
python CORE/scripts/story.py outline --work tf --genre testimonial-film
# result hook first, then the story in order; read tf/story_flags.md
python CORE/scripts/story.py reorder --work tf --order "s14,s2-s13,s15-s18" --genre testimonial-film --format 16:9 --out tf/order.json
python CORE/scripts/run.py - --timeline tf/order.json --style SKILLS/testimonial-card --genre testimonial-film --format 16:9 --broll broll/ --music bed.wav --work tf2 --stop-after broll
python CORE/scripts/render.py tf2/plan.json --frames 1,10,30,60,85
python CORE/scripts/render.py tf2/plan.json
python CORE/scripts/finish.py tf2/plan.json --platform youtube
Long interview with many retakes: selects.py --work sel --input customer.mp4, edit sel/paper_edit.md,
selects.py --work sel --apply --format 16:9, then run.py - --timeline sel/selects_spec.json ....
Vertical cut: same order spec with --format 9:16 --duration 30 in a new work folder, or
track.py reframe customer.mp4 --out cust_916.mp4 --format 9:16 first.
Refine the plan
- Hook text (muted viewers): a
textelement in the first 2.5 s with the result in their words, quoted exactly. - Lower third at their first line:
{"type": "lowerthird", "name": "Ravi Kumar", "role": "Ops Lead, Acme Foods", "start": 2.8, "end": 7}. - Result numbers:
{"type": "counter", "from": 0, "to": 38, "suffix": "%", "label": "fewer late orders", "start": 52.1, "end": 56}only for numbers the customer says or the client documents. Check make_plan's auto counters against what is said. - Quote card:
{"type": "quote", "text": "We got our weekends back.", "author": "Ravi Kumar", "role": "Acme Foods"}, word for word. - Disclosure (when compensated):
{"type": "chip", "text": "Paid partnership", "start": 0, "end": 4}or an end card line. - Captions:
{"type": "captions", "preset": "clean"}; fix names and product terms inwords(keep times). - Colour/audio:
"color": {"look": "documentary-natural", "intensity": 0.5},"music": {"src": "bed.wav", "under": 22, "duck": "speech-aware", "fitTo": "beats"},"roomTone": true. python CORE/scripts/editor.py review tf2/plan.json --genre testimonial-film; fix uncovered jump cuts it lists.- B-roll videos (
cutaways) show in the--framesstills; check each lands on the right line.
Deliver
python CORE/scripts/deliver.py tf2/plan.json --to youtube,linkedin --out-dir tf2/deliver
python CORE/scripts/deliver.py tf30/plan.json --to reels,shorts --out-dir tf30/deliver
python CORE/scripts/qc.py tf2/deliver/final_youtube.mp4 --plan tf2/plan.json --platform youtube --strict
Hand over the film, the 30 s cut, verticals, captions.srt, 3 thumbnails (thumb_*.png) and the change log of any
reordered sentences for the customer's approval.
Check before delivering
- Every sentence is whole and in context;
story_flags.mdhas no unresolved HIGH flags. - Name, title, company and numbers match the customer's approval; disclosure present if they were compensated.
- No visible jump cut; voice clean but natural;
qc.pypasses loudness and caption checks. - The customer (or the client on their behalf) has approved the final cut.
Example requests
- "Make a 90-second case study from this interview with our client, best line first."
- "Cut a 30-second vertical testimonial for Instagram with captions."
- "Add their results as numbers on screen: 38% fewer late orders, 6 hours saved a week."
Files (claude-video-editing-skills)
-
SKILL.md 7.4 KB
--- name: testimonial-film description: "Customer testimonial editor for Claude: turns a real customer interview into a 60-150 s case-study film (result hook, problem, solution, numbers, recommendation) plus a 30 s cut and vertical clips, with the best soundbite first and every jump cut covered. Use when the user says 'edit this customer testimonial', 'make a case study video', 'client review video', 'success story', 'social proof video', 'cut the interview with our customer', or hands over a customer interview and product B-roll. Real customers, real words, real order only. Runs on video-studio-core (transcribe, selects, story, B-roll, mix, deliver, QC)." license: MIT --- # Testimonial Film A real customer, in their own words, saying what changed: credible first, polished second. - **Best for:** customer case studies, client reviews, agency results, B2B social proof, course student stories - **Inputs:** customer interview (one or two angles), B-roll of the customer using the product, result screenshots, logo - **Default skin:** `testimonial-card` · alternatives `split-proof`, `quote-card-reel`, `frame-studio` - **Formats:** 16:9 master, 9:16 and 4:5 cutdowns - **Durations:** film 60-150 s, cutdown 30 s, Shorts 15-45 s - **Genre preset:** `--genre testimonial-film` ## Before you start - Needs **video-studio-core** in the same skills folder. `CORE` = that folder, `SKILLS` = the skills folder. - First run: `python CORE/scripts/check_env.py`. - **Ethics (non-negotiable):** the speaker is a real customer, the words are theirs, and their meaning is unchanged. No actors presented as customers, no invented numbers, no `stars` rating unless the user supplies the real review. If the customer was paid, got a discount or free product, add a disclosure card (FTC / ASCI). Get the customer's sign-off on the final cut. - Ask for: name, job title, company (spelled exactly), the result numbers and where they come from, logo permission. ## How a pro cuts this **Structure (% of length)** | beat | % | what goes here | | --- | --- | --- | | result hook | 10 | the single best soundbite: the outcome in their words | | problem before | 25 | what life was like, the cost of the problem | | solution | 30 | how they found and used the product, specific moments | | result / numbers | 25 | concrete results, shown with counters or screenshots | | recommendation + CTA | 10 | "I'd tell anyone who..." then logo and CTA | **Pacing.** ASL 3-6 s, visual change at least every 5 s. Hold the face on emotional or credibility lines; B-roll the explanations. **Cut rules.** - Paper edit from the transcript; the hook may move up from later in the interview (log it), but every sentence stays whole. `story.py reorder` flags partial splices (HIGH) and back-references (MED): resolve every flag. - Cover every join: B-roll >= 0.5 s each side, second angle, or a >= 1.2x punch-in (`speaker.zooms`). - Light filler removal; keep natural hesitations that make them sound real. Keep the breath before the key line. - Never use a reaction or B-roll that changes the meaning (Kuleshov). **Transition grammar.** Cuts and J/L cuts; one dissolve at most into the CTA. No flashy transitions. **Music and sound.** Bed ~-24 dB under speech, ducked (`duck: "speech-aware"`), lifted at the recommendation. Room tone on every join. SFX low (4/min): a soft tick on a counter. **Colour.** Natural and trustworthy: `documentary-natural` or `portra` at 0.5. Brief says "punchy": `clean-commercial` 0.6. **Captions.** Two-line sentence-case subtitles low in frame (`"preset": "clean"`), <= 17 cps; burned captions on verticals. **What the client judges.** Credibility, the customer sounding natural, a concrete result. **Common mistakes.** Starting with "Hi, my name is"; numbers nobody can verify; over-polished audio; obvious jump cuts; music louder than the customer; a quote stitched from two answers. ## Make the edit ```bash python CORE/scripts/run.py customer.mp4 --style SKILLS/testimonial-card --genre testimonial-film --format 16:9 --smart-cuts --retakes --logo logo.png --work tf --stop-after plan python CORE/scripts/story.py outline --work tf --genre testimonial-film # result hook first, then the story in order; read tf/story_flags.md python CORE/scripts/story.py reorder --work tf --order "s14,s2-s13,s15-s18" --genre testimonial-film --format 16:9 --out tf/order.json python CORE/scripts/run.py - --timeline tf/order.json --style SKILLS/testimonial-card --genre testimonial-film --format 16:9 --broll broll/ --music bed.wav --work tf2 --stop-after broll python CORE/scripts/render.py tf2/plan.json --frames 1,10,30,60,85 python CORE/scripts/render.py tf2/plan.json python CORE/scripts/finish.py tf2/plan.json --platform youtube ``` Long interview with many retakes: `selects.py --work sel --input customer.mp4`, edit `sel/paper_edit.md`, `selects.py --work sel --apply --format 16:9`, then `run.py - --timeline sel/selects_spec.json ...`. Vertical cut: same order spec with `--format 9:16 --duration 30` in a new work folder, or `track.py reframe customer.mp4 --out cust_916.mp4 --format 9:16` first. ## Refine the plan - **Hook text (muted viewers):** a `text` element in the first 2.5 s with the result in their words, quoted exactly. - **Lower third** at their first line: `{"type": "lowerthird", "name": "Ravi Kumar", "role": "Ops Lead, Acme Foods", "start": 2.8, "end": 7}`. - **Result numbers:** `{"type": "counter", "from": 0, "to": 38, "suffix": "%", "label": "fewer late orders", "start": 52.1, "end": 56}` only for numbers the customer says or the client documents. Check make_plan's auto counters against what is said. - **Quote card:** `{"type": "quote", "text": "We got our weekends back.", "author": "Ravi Kumar", "role": "Acme Foods"}`, word for word. - **Disclosure** (when compensated): `{"type": "chip", "text": "Paid partnership", "start": 0, "end": 4}` or an end card line. - **Captions:** `{"type": "captions", "preset": "clean"}`; fix names and product terms in `words` (keep times). - **Colour/audio:** `"color": {"look": "documentary-natural", "intensity": 0.5}`, `"music": {"src": "bed.wav", "under": 22, "duck": "speech-aware", "fitTo": "beats"}`, `"roomTone": true`. - `python CORE/scripts/editor.py review tf2/plan.json --genre testimonial-film`; fix uncovered jump cuts it lists. - B-roll videos (`cutaways`) show in the `--frames` stills; check each lands on the right line. ## Deliver ```bash python CORE/scripts/deliver.py tf2/plan.json --to youtube,linkedin --out-dir tf2/deliver python CORE/scripts/deliver.py tf30/plan.json --to reels,shorts --out-dir tf30/deliver python CORE/scripts/qc.py tf2/deliver/final_youtube.mp4 --plan tf2/plan.json --platform youtube --strict ``` Hand over the film, the 30 s cut, verticals, `captions.srt`, 3 thumbnails (`thumb_*.png`) and the change log of any reordered sentences for the customer's approval. ## Check before delivering - Every sentence is whole and in context; `story_flags.md` has no unresolved HIGH flags. - Name, title, company and numbers match the customer's approval; disclosure present if they were compensated. - No visible jump cut; voice clean but natural; `qc.py` passes loudness and caption checks. - The customer (or the client on their behalf) has approved the final cut. ## Example requests - "Make a 90-second case study from this interview with our client, best line first." - "Cut a 30-second vertical testimonial for Instagram with captions." - "Add their results as numbers on screen: 38% fewer late orders, 6 hours saved a week."
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.