Four studio tools,
rebuilt to a professional standard.
One showcase closes the series.

The eight columns — exactly one row of the Phase 1 log

• 01Title — plus two alternates, so the angle can be re-shot under a different headline
• 02Hook (first 3 seconds) — one of five interchangeable hook styles
• 03Script — Hook / Body / CTA structure, with its real word count shown
• 04Caption — platform-aware sign-off for TikTok, Reels or Shorts
• 05Hashtags — niche + topic set, deliberate casing kept (#ChatGPT stays #ChatGPT)
• 06Veo 3 Prompt — subject, per-angle scene, camera, lighting, platform closing line
• 07Negative Prompt — blocks deformed hands, text overlays, watermarks, CGI look
• 08Duration — target, real word count vs. budget, estimated read time

Controls that actually change the output

Six angles — step-by-step tutorial, common mistakes, comparison, surprising facts, personal story, quick tips — each with its own point bank, opener and CTA, so the angle changes the argument, not just the title. Five hook styles (curiosity question, bold claim, mistake to avoid, number / specific, before–after) can front any angle. Three tones — Friendly teacher, Bold & direct, Calm expert — swap the opener and CTA register while the points stay identical. Three platforms and three durations (30 / 45 / 60 seconds) complete the brief; script language is fixed to English.

Timing formula: spoken English at a natural creator pace runs at roughly 140 words per minute, so the budget is duration × 140 ÷ 60 — ±70 words at 30s, ±105 at 45s, ±140 at 60s. Every angle keeps three cuts of its points — short, medium, full — and the generator picks the cut that fits (2 short points at 30s, 3 medium at 45s, 4 full at 60s).

Verified sample package — rebuilt and re-measured for this page

At build time the generator was exercised headlessly across all 54 combinations (6 angles × 3 tones × 3 durations) and every package came out complete — eight columns plus two title alternates. An early draft overshot its word budget by +60–70% at 30 seconds; the three-cut point system brought the worst deviation down to +11%. The sample above sits at 108 words against a 105-word budget (+3 words).

The 17-field package every card carries

Title · Niche / Topic · Target Audience · Hook · Narration Script (HOOK / BODY / CTA) · Script Shots · Visual Prompts · Veo Prompts · Negative Prompt · VO Notes · Subtitle Text · Caption · Hashtags · Thumbnail Text · Publish Notes / Link · Duration (30 / 45 / 60 seconds) · Notes. Fifteen text fields plus duration and notes — the same package shape from the LW-01 generator, extended with the fields production actually needs (shots, subtitles, VO direction, thumbnail text, publish notes).

The five QA gates — a field item passes when its field is filled; a manual item passes when the Boss ticks it

The seeded EXAMPLE card — a complete package, held by its two manual checks on purpose

Every field of the example is filled, but its two manual checks — the VO take approval and the Boss's final QA approval — are deliberately left unticked, so the gate is visible in action: the card cannot advance past Voice Over or QA until a human listens and approves.

Live board snapshot — read from the board itself for this page

Read on Friday, 6 November 2026 via the board's own listing: exactly one card — the seeded EXAMPLE above — sitting in Idea. Its Idea, Script and Edit gates pass on filled fields; Voice Over and QA stay open on the two unticked manual approvals, so publish-ready is honestly 0. This is live working data, not a series statistic: the counts change the moment a card is added, ticked, or moved, and this snapshot does not update itself.

Controls, exactly as built

Scroll speed slider 15–220 px/s (default 70 px/s) · text size slider 26–76 px (default 44 px) · Space plays / pauses outside the script box, R returns to the start · mirror toggle flips the text horizontally · the editor, counter and benchmark table all update live as the script changes — there is no save button. Everything runs in the browser; the script is never sent anywhere.

Tempos: Relaxed at ±110 words/min · Normal at ±140 words/min · Fast at ±170 words/min. Benchmarks (words that fit each target): 30 seconds → ±55 / ±70 / ±85 words · 45 seconds → ±83 / ±105 / ±128 · 60 seconds → ±110 / ±140 / ±170. The table highlights the row your current script lands closest to at Normal pace.

Sample script — counted by the rebuilt tool's own rule

Measured for this page with the tool's own counting method (markers stripped, then whitespace-split): 134 spoken words → Relaxed ±1 min 13 sec (110 wpm) · Normal ±57 sec (140 wpm) · Fast ±47 sec (170 wpm). This supersedes the figure quoted on the previous edition of this showcase, which belonged to the old build's different sample script.

The six-point VO guide: 01 Win the first 3 seconds · 02 Pause between points · 03 Lift your pitch on questions · 04 Keep the energy level · 05 Record where echo dies · 06 Check your levels first (a 5-second test take, checked on earphones).

Platform specifications — shown live in the tool before you generate

Inputs: video title (required — an empty title is blocked with an error, never a silent empty package) · platform · niche · an optional hand-written hook line · one of five CTAs — follow, comment, save, share, or check the profile. The title field counts characters live as you type, against the selected platform's limit.

Package Quality Check — five automatic checks on every package

1. Title fits the platform limit (reported as characters used / limit) · 2. Hashtag count inside the benchmark for that platform · 3. Hook source — credited when hand-written, flagged “consider a hand-written hook” when auto-generated from the title · 4. Topic keyword found for the pinned comment · 5. Caption is substantial enough to post as-is (at least 120 characters). Each check reports Pass or Check with its measured detail — nothing is waved through.

Pre-publish checklist (5 items, with a “n of 5 ready” progress line): cover text / hook readable within the first 3 seconds · caption pasted into the description field · hashtags pasted (caption or comment) · pinned comment written and pinned · posting time selected. Posting-time guidance shown in the tool is general — Indonesian audiences are generally busiest at lunch, 12:00–13:00 WIB, and in the evening, 18:00–21:00 WIB — and the tool says plainly that an account's real best time only shows in its own statistics.

Honest limit, stated on the tool itself: it only prepares the package. Posting still happens by hand (or through the Boss's own workflow) on each platform — the tool is connected to none of them and cannot post on anyone's behalf.