IAmCosmin-YT/emotion-factory-devlog-shorts

Devlog & High-Retention Shorts Editing Skill (81-Video Frame-by-Frame Forensic Edition)

A complete production manual and Python/OpenCV/PIL 60fps frame-by-frame compositing system for creating viral YouTube Shorts, TikTok devlogs, and gaming promos. Derived from a frame-by-frame forensic inspection of 81 videos (162 contact & hook sheets) across Vedinad (all 28 Shorts + 2 long-form videos), Danidev (15 devlogs, Dani2 videos, and DaniClips shorts), Wishbone Games (10 Shorts and devlogs), c dev (14 Shorts and devlogs), SSundee (5 intro hooks), and 7 viral reference Shorts (tt_wynge, ref5, story_video, speed_main, speed_rivals, speed_toh, speed_flop).

---

1. The 7 Non-Negotiable Laws (From Frame-by-Frame Evidence)

1. NO Persistent Top Header Banner:

   • Never pin a wide static rectangular header banner across the top of the video. Across all 15 Danidev videos, 10 Wishbone videos, 14 c dev videos, 5 SSundee videos, and 7 reference shorts (tt_wynge, ref5, story_video, speed_*), zero use a persistent top header banner.
   • Even in Vedinad's shorts, many (PvmyoTRTKho, qEOSXthsSUU, etc.) have no top text at all, or use a compact, temporary native TikTok text sticker that disappears or changes by act (0–4s hook -> disappears during B-roll -> temporary "Wait for it..." -> disappears). Never cover top gameplay HUDs with a permanent banner.

2. True Frame-by-Frame Hook Architecture (0.00s – 2.50s):

   • Archetype A — Animated Word-by-Word Comment/Discord Card on Clean Grid or Dark/Blurred Backdrop (Vedinad / Dani / c dev):
     • Opens on a clean diagonal-scrolling blue tile grid (#1A66CC to #2B82F6), dark slate grid (#12131A), or heavily blurred gameplay (GaussianBlur(25)).
     • A 3D-tilted YouTube comment card or dark Discord message (#313338) starts at 1.0x scale (-3.5° to -5° tilt). As the comment text types on word-by-word in sync with the voiceover (0.00s–1.60s), the virtual camera zooms in (1.0x -> 2.0x) and pans horizontally from left to right tracking each word.
     • At 1.60s–2.10s, the camera snaps back out to 1.0x (flipping tilt to +3.5°) on a 1–2 word narrator reaction ("UH OH", "OKAY..", "BRO...", "CHALLENGE ACCEPTED"), and vanishes completely via hard cut at 2.00s–2.50s.
   • Archetype B — High-Impact Motion-Graphics / Counter Slam Hook (tt_wynge / c dev rcS9EkA_pco / Wishbone):
     • Opens immediately on a full-screen 3D-tilted UI card, draining/ticking stat bar ($8,356 -> $0 or 0 -> 278,824), or 1-word center-screen neon glow slams ("This" -> "Roblox" -> "Game") with RGB chromatic aberration and screen shake.

3. Medium Switching & B-Roll Cutaways Every 1.0s – 1.7s (35–55 Cuts/Min):

   • Raw gameplay alone—even with zoom cuts—causes viewer fatigue if not broken up by full-screen graphic pattern interrupts. Every 4–6 seconds of tight gameplay must be punctuated by one of these 5 signature devlog cutaways (lasting 0.6s–1.8s):
     1. Black-Screen Deadpan Text Slam (#000000): Pure black screen with centered text slamming in ("but it gets WORSE", "IT GETS EVEN CRAZIER", "SO YEAH", "WAIT FOR IT...", or lowercase deadpan "rip bean").
     2. Black-Screen Numbered Dev Checklist: Typing out a 3–4 item list on pure black ("1. build emotion factory", "2. hire grumpy workers", "3. 10x lightning storm", "4. profit??").
     3. 3D Perspective-Tilted Code / Studio Inspector Card: A dark IDE (#0D1117) or Inspector card tilted in 3D perspective (cv2.warpPerspective) with syntax-highlighted comedic code (cube.Value *= 10; // balanced lol) or speed-ramped typing (400%–800%).
     4. Animated Number Counter / Multiplier Card: A dark blurred card where a massive counter ticks up rapidly (1x -> 10x -> 1,000x or $0 -> $1,000,000,000) with a glowing progress bar.
     5. Stacked Rarity / Tier Ladder Card: Vertical stack of color-coded horizontal pill bars (HAPPY $5 -> GRUMPY $75 -> RAGE $400 -> SHOCKED 10x -> ASCENDED $10,000) popping in sequentially.

4. Clean, Tight Gameplay Cropping (Zero Cluttered Menus):

   • When cropping 16:9 gameplay to 9:16 (1080x1920), never include accidental settings popups, save-slot dialogs, or tiny unreadable sidebars.
   • Crop tightly (scale 1.35x–1.85x) onto the active focal point: the character, conveyor items, physics collisions, boss lasers, or floating damage/cash numbers. Use subtle top/bottom dark gradient vignettes (alpha 0.0 -> 0.55) so kinetic typography has 100% contrast.

5. Giant High-Contrast Kinetic Typography (95pt – 160pt):

   • Never use tiny 50pt–60pt captions buried at the bottom of the screen.
   • Standard Gameplay Position: Lower-middle safe zone (Y = 1280 to 1380, 67%–72% from top)—sitting cleanly below the center character action and well above the bottom UI rail (Y > 1550).
   • Full-Screen Card / Slam Position: Dead center (Y = 860 to 980, 45%–51% from top).
   • Styling: Heavy geometric sans-serif (Segoe UI Black, Montserrat Black, Arial Black), ALL CAPS, 1 to 3 words max (or 2-line stacked punch compounds like "EMOTION\nTYCOON", "10X SHOCK\nVALUE"), 12px–16px solid #000000 stroke + deep drop shadow + optional colored neon outer glow (Red, Gold, Neon Green, Cyan).
   • Spring & Blur Entry: Each word chunk pops in with a 4-frame spring overshoot (70% -> 116% -> 100%) and optional 1-frame horizontal directional motion blur on fast transitions.

6. Crisp, Natural Browser Voiceover (Zero Robotic Edge-TTS):

   • Never use flat, metallic CLI synthesizers (edge-tts or OS SAPI).
   • Generate voiceover through an incorporated Chromium browser via Playwright using expressive neural voice engines (such as https://www.openai.fm with custom voice-acting prompts or browser ElevenLabs when unblocked), then run purge_silence.py to strip all pauses > 80ms and extract exact frame-accurate word timestamps with Whisper (words.py).

7. The Seamless Infinite Loop (Verbal + Visual):

   • The final 1.0s of the video must never fade out or say "Thanks for watching, subscribe!". Instead, end on an abrupt comedic cut or incomplete grammatical clause ("...which is why people keep saying...") that resolves directly into Frame 0 ("...you CAN'T make a factory that farms literal emotions!").

---

2. End-to-End Production Pipeline

Step 1: Browser Voiceover Generation & Word Timestamping

1. Write a tight 30–38 second script (95 to 125 words at 185–210 WPM) structured into 5 rapid acts with an infinite loop ending.
2. Generate crisp, expressive audio via Playwright Chrome (scripts/generate_browser_tts.py), purge dead air > 80ms (scripts/purge_silence.py), and extract exact word-level timestamps using words.py.

Step 2: Frame-by-Frame 60fps OpenCV + PIL Compositor

Instead of relying on static FFmpeg overlay filters, render the video at 1080x1920 @ 60fps using a frame-by-frame Python compositor (cv2 + PIL + numpy) that computes exact per-frame animation math (ease_out_back, ease_in_out_cubic, 3D perspective warps, camera shake, RGB glitch splits, and particle overlays):

• 0.00s – 2.20s (Act 1 — Comment Zoom Hook): Animated blue/dark grid background + 3D-tilted comment card typing word-by-word while zooming 1.0x -> 1.85x and panning across words, snapping back on the narrator's reaction.
• 2.20s – 8.50s (Act 2 — Premise & Core Loop): Tight pan-and-scan gameplay cuts (1.1s–1.5s each) interleaved with a 3D-tilted Luau/C# script card and a black-screen numbered dev checklist.
• 8.50s – 21.0s (Act 3 — Mechanic Escalation & Synergies): Fast gameplay showcases of upgrades/workers/synergies punctuated by a black-screen deadpan slam ("BUT IT GETS WORSE") and a 5-tier stacked Rarity/Emotion Ladder card.
• 21.0s – 31.0s (Act 4 — Weather/Boss Climax & Animated Counter): High-intensity event gameplay (e.g., Thunderstorm 10x lightning strikes, Blood Moon crimson cubes, Hydraulic Compactor) paired with a ticking Multiplier/Cash Counter card ($1 -> $1,000,000,000) and full-screen impact flashes.
• 31.0s – 34.5s (Act 5 — CTA & Seamless Loop Outro): Crisp game title/play CTA card transitioning mid-sentence back into Frame 0.

Step 3: Hyper-Compressed Audio Mix (-14 LUFS, LRA 3.1 LU)

Mix the silence-purged voiceover (0 dB) with ducked upbeat background music (-20 dB under voice) and real royalty-free SFX (Pixabay / MyInstants / Kenney—never AI-generated noise):

• Keyboard typing clicks synced to every word appearing on the hook comment card and checklist card.
• Crisp whooshes on every camera punch-in and 3D card flip.
• Pop / wood clicks on keyword subtitle appearances.
• Bass boom / Vine boom on black-screen deadpan slams and climax reveals.
• Cash chimes / lightning thunder on counter finishes and multiplier hits.

---

3. Detailed Forensic Reference Manuals

• Creator Breakdowns & 81-Video Forensic Profiles: Complete per-creator rules distilled from all 30 Vedinad videos, 15 Danidev videos, 10 Wishbone Games videos, 14 c dev videos, 5 SSundee videos, and 7 reference Shorts.
• Hook Architecture & Retention Loops: Exact frame-by-frame math for Vedinad/Dani comment zoom hooks, tt_wynge/c dev motion-graphics hooks, and infinite loop scripting.
• Kinetic Subtitles & Typography Specification: Font sizing (95pt–160pt), 1–2 line stacking, spring overshoot curves, directional blur smears, and semantic glow/color palettes.
• Pacing, Cutaways & B-Roll Engineering: ASL metrics across all 81 videos, black-screen text slams, 3D tilted code cards, rarity ladders, and 9:16 pan-and-scan framing rules.
• Sound Design, Browser TTS & Audio Mastering: Playwright browser TTS generation, silence purging, SFX sync map, and -14 LUFS mastering chain.

---

4. Quality Gate & Visual Verification Checklist

Before delivering any rendered Short, generate a 60-frame contact sheet + 24-frame 4fps hook sheet and inspect both with view_file:

• No Persistent Top Header Pill: Confirm there is NO permanent white/black banner pinned across the top of the video.
• Frame 0–60 Hook Movement: Does the opening comment card or motion-graphics hook actively type on word-by-word and zoom/pan across 0.0s–2.2s, then vanish completely?
• Clean Gameplay Crops: Are 100% of gameplay shots free of accidental SETTINGS, SAVES, or cluttered menu popups?
• Visual Medium Variety: Does the video switch between tight gameplay and full-screen graphic cutaways (black-screen slams, 3D code cards, animated counters, tier ladders) at least every 4–6 seconds?
• Oversized Legible Typography: Are subtitles 95pt–150pt with thick black outlines/glows, positioned cleanly in the safe zone (Y = 1280–1380 or center Y = 880–960 on graphic cards) without overlapping bottom UI buttons?
• Voiceover Crispness: Was the voiceover generated via browser neural TTS (not robotic CLI edge-tts) with all pauses > 80ms removed?