vaibhav-template

name: vaibhav-template description: Turn any script into a Vaibhav Sisinty-style talking-head video for Graeham Watts — the visual formula reverse-engineered from @vaibhavsisinty's 1.7M-follower Instagram reels. Fires on "run this script through the template I love", "make this Vaibhav-style", "use my video template", "convert to Vaibhav format", "add the warm desk aesthetic", "put this in my rotation", or any request to produce a video where Graeham's face appears with the burned-in visual system (locked talking-head, fast front-loaded cuts, serif-italic captions, color-highlighted keywords, emoji pop-ins). Encodes Graeham's 5 pre-generated warm-lit avatar looks (rotate across videos for variety without breaking brand consistency), the 5-mode composition system, cut pacing arc, and typography specs. Hands script + look off to heygen-video for rendering and higgsfield-video for B-roll. Use this instead of freestyling a HeyGen render when the user wants the Vaibhav aesthetic.

Vaibhav-Style Video Template — Graeham Watts

A reusable visual formula reverse-engineered from @vaibhavsisinty (1.7M followers, Instagram Reels). This skill doesn't make a video from scratch — it takes a script or topic and produces a shot plan, look choice, and caption spec that matches the Vaibhav aesthetic, then hands it to heygen-video for rendering and higgsfield-video for B-roll.

When this fires

• "Run this script through the template I love"
• "Make this Vaibhav-style"
• "Use my video template on this"
• "Convert this script to the warm-desk look"
• "Add this to my content rotation"
• "Put this in the Vaibhav format"

If the user has a topic but no script yet, chain with content-creation-engine first — that writes the script (it absorbed video-script-creation-engine in April 2026), then hand it back to this skill.

The reality

This is a STYLE skill, not a rendering skill. The actual video comes out of heygen-video. This skill's job is to make sure every HeyGen video Graeham publishes has the same visual grammar — so viewers recognize his content within 2 seconds regardless of topic.

Vaibhav's consistency is not an accident. His reels are recognizable within one frame because he locked in exactly five things: talking-head framing, cut pacing arc, typography, color grade, and caption placement. Those five things are this skill's entire contract.

Visual system — the 5 composition modes

Every Vaibhav-style video is built from these 5 frames remixed. Don't invent new ones mid-video.

Mode 1 — Hook composite (seconds 0–3 ONLY)

• Top 60% of frame: full-bleed B-roll of the HOOK subject (Sam Altman, a listing, a map, a headline)
• Bottom 40% of frame: LOCKED talking-head of Graeham, same position every video
• Caption on split line: serif italic subject name top, sans-serif supporting clause underneath
• Source files: Top needs Higgsfield B-roll; bottom uses one of Graeham's 5 warm-desk looks
• Used for: Opening hook, re-engagement moment mid-video if attention drops

Mode 2 — Full-bleed talking head

• Full frame: Graeham's face, chest-up
• Optional overlay: red or orange color wash at 30% opacity (section transition)
• Caption: large serif italic center-frame
• Used for: Section transitions, high-emphasis statements, the "let me show you" pivot

Mode 3 — Full-bleed B-roll

• Full frame: environmental B-roll (street shot, listing exterior, neighborhood)
• Caption: small white-on-dark translucent pill, center or bottom-third
• Used for: Market data, location grounding, visual establishment

Mode 4 — Screenshot card (PiP)

• Background: solid dark/black
• Top 60%: product screenshot, document, or tool output (prompt card, listing MLS shot, Redfin page)
• Bottom: small inset of Graeham + big acid-green list number (#01, #02) + 3-line caption
• Used for: "Here's how to do it" content, tool walkthroughs, step-by-step

Mode 5 — Section header

• Background: current warm-desk look, faded/darkened
• Watermark center-back: PropertyIQ logo or Graeham Watts Investment Properties mark at 20% opacity, large
• Top overlay: ##. TITLE CASE in letter-spaced sans-serif
• Used for: Opening each numbered section in a listicle format

Cut rhythm — the documented arc

From the reel analysis (80s, 52 cuts, scene-detection verified):

Key rule: front-load 40% of your cuts in the first 10% of the runtime. This is what makes Vaibhav's style feel "fast" even though 80% of the video is moderately paced. Miss this and the video will feel slow no matter how good the rest is.

Translation to production: for a 60-second video, plan ~16 cuts in the first 6 seconds, then ~20 cuts over the remaining 54s.

Typography — the distinctive signature

His fonts are what make this aesthetic look premium instead of generic-viral. Most reels editors use bold sans-serif. He does the opposite.

Color-highlighted keywords: 1–3 emphasis words per caption get acid-green or warm-yellow highlight boxes behind the text. Only highlight nouns and verbs that carry meaning — never articles, connectors, or filler.

Color & grade

• Base palette: warm tungsten orange (3200K) + cool dusk blue (5600K), simultaneously.
• Grading move: subject is warmly lit; background is cool and blurred. This is Vaibhav's single most important color rule.
• Section washes: red (30% opacity, sharp transition) and gold (20% opacity, gentle transition) — use to mark scene shifts, not every cut.
• Never: flat white balance, cool-only grade, or gray grade. The warm/cool split IS the look.

Graeham's 5 looks — the rotation system

Five avatar looks, each shares the "warm-lit desk with subject facing camera head-on, desk horizontal across bottom of frame" foundation but varies outfit + environment. Rotate across videos so viewers see variety but instant brand recognition.

Rotation rule: match look to content intent, not randomly.

• Market update → warm_desk_navy
• Bold take / podcast-style → podcast_studio
• Neighborhood / lifestyle → loft_window
• Seller meeting, listing intro, "I just sold this for $X" → corporate_office
• "Here's how a contract works" / educational → modern_studio

Don't rotate mid-video. One look per video. Switching looks inside a single reel breaks identity.

The skill's workflow

Step 1: Understand the script's job

Before doing anything, identify:

• Content type: market update / bold take / lifestyle / seller-facing / educational
• Target runtime: 30s / 60s / 90s — this drives cut planning
• Hook payload: what ONE specific thing in the first 3 seconds stops the scroll
• CTA: what action at the end

Ask the user if any of these are unclear. Don't assume.

Step 2: Pick the look

Match the content type to the look table above. State the pick with reasoning. Example:

"This is a market-data video about EPA pricing. I'd use warm_desk_navy — it's the everyday default and the data-driven tone doesn't need the gravitas of corporate_office or the studio polish of modern_studio. Good?"

Confirm before proceeding. Don't assume the user wants the default.

Step 3: Build the shot plan

Produce a table mapping each line/beat of the script to a composition mode + timing + caption spec. Example for a hook:

Time    Mode      Caption                              B-roll needed
0–1s    Mode 1    "Sam Altman" (italic, top)           portrait of person/subject
                  "just killed the entire" (bottom)    
1–2s    Mode 1    continues, crossfade caption         same / slight zoom
2–3s    Mode 2    "Image Gen Industry"                 Graeham full-bleed
                  (italic, yellow highlight on "Image")
3–4s    Mode 1    "with one big launch"                4K industrial image
4–5s    Mode 5    "01. PRECISE TEXT"                   Graeham warm_desk_navy + PropertyIQ logo fade

Front-load cuts per the rhythm arc. Fill in the Body and Climax with Modes 3 and 4. End in Mode 2 with a clean CTA.

Step 4: List the B-roll Graeham needs

For each beat that requires a B-roll clip (Modes 1, 3, and sometimes 4), produce an explicit Higgsfield prompt. Hand these to the higgsfield-video skill when the user is ready to shoot them.

Example:

B-roll #1 (beat at 0–3s):
  Orientation: 16:9 (landscape)
  Prompt: "4K ultra-detailed portrait of tech CEO in black turtleneck, 
          multiple facial expressions composited in a grid, dark 
          dramatic studio lighting, shot on Kodak Portra 400, Douglas 
          Friedman editorial style. Horizontal 16:9."
  Duration: 4s
  Motion: slow push-in

Step 5: Hand off to heygen-video

Once look + script are locked, call the heygen-video skill with:

python3 /path/to/heygen-video/scripts/create.py \
  --script "..." \
  --look <chosen_look> \
  --aspect 16:9 \
  --title "<content type> - <date>"

Aspect: All 5 Vaibhav looks are landscape-native (16:9). Graeham edits to portrait (9:16) in post. Always render at 16:9 from HeyGen.

Step 6: Build the caption + typography sheet

Deliver a simple text file listing each caption beat, font, color, highlight words, and timing. Graeham's editor (or CapCut operator) uses this to burn in captions matching the Vaibhav typography spec.

What this skill does NOT do

• ❌ Does not write the original script — use content-creation-engine or handle that separately
• ❌ Does not render video — always hands off to heygen-video
• ❌ Does not generate B-roll — produces prompts for higgsfield-video to execute
• ❌ Does not edit/composite final video — Graeham or his editor does this in CapCut / Premiere / After Effects using the shot plan + caption spec
• ❌ Does not let Graeham "freestyle" mid-video — one look, one grade, one typography system per video

Why consistency matters (and why to resist "just this once" deviations)

Vaibhav's visual system looks simple because he never breaks it. 3,000+ posts at 1.7M followers — the same five composition modes, the same typography, the same warm/cool grade. Viewers recognize his content before the first word plays. That's the brand moat.

If Graeham mixes looks mid-video, or swaps fonts for "variety", or front-loads 5 cuts instead of 20, he'll get a video that's fine but doesn't read as his. That's the difference between "another real estate agent posting on IG" and "the Peninsula guy who does those cinematic reels."

Strict adherence to this template IS the strategy.

Quick reference card

Input:    script + content_type
Output:   shot plan + look choice + caption spec + B-roll prompts + HeyGen invocation

LOOK DECISION TREE
  market-data / everyday          → warm_desk_navy
  bold opinion / hot take          → podcast_studio
  lifestyle / neighborhood         → loft_window
  seller-facing / listing intro    → corporate_office
  educational / contract breakdown → modern_studio

CUT RHYTHM
  First 10% of runtime:  40% of all cuts  (≈0.5s shots)
  10–38%:                15% of all cuts  (≈2.5s shots)
  38–75%:                27% of all cuts  (≈2.0s shots)
  75–88%:                13% of all cuts  (≈1.4s shots)
  Last 12%:              6% of all cuts   (≈3.0s shots)

TYPOGRAPHY
  Subject/emphasis:  Playfair Display Italic, white
  Secondary:         DM Sans Regular, white
  List numbers:      DM Sans Bold, acid-green #BFFF00
  Section titles:    DM Sans Medium, letter-spaced 8%, UPPERCASE
  Captions:          Inter Semi-bold on dark translucent pill

GRADE
  Warm face (3200K) + cool background (5600K) = the signature
  Section washes: red 30% (sharp), gold 20% (gentle)

COMPOSITION MODES (use only these 5)
  Mode 1  Hook composite (60% B-roll top / 40% locked Graeham bottom)
  Mode 2  Full-bleed talking head
  Mode 3  Full-bleed B-roll
  Mode 4  Screenshot PiP card
  Mode 5  Section header w/ logo watermark

ASPECT: render HeyGen at 16:9. Edit to 9:16 in post.

Session history

Built 2026-04-23 with Graeham over a long working session. Reverse-engineered from an 80-second Vaibhav Sisinty reel (160 extracted frames, 52 scene-detected cuts, pacing arc measured per section). All 5 looks generated via Higgsfield Nano Banana Pro with identity-matched reference image, verified against Graeham's own selfie (IMG_0520.JPG). Anti-cleft prompt formula documented in references/prompt_formula.md — use this when regenerating or adding a 6th look.