The Avatar Takes Over the Daily Shoot

A busy professional is asked to make more content, then appears to be filming several examples at once. A visually matched second version of that professional enters, explains the product's benefit through conversation, and takes over the final direct-to-camera promise. The product reveal is a character reveal at about 0:28, after the audience has seen the work it solves.

Why it works

• The first question names a concrete workload at 0:00, and three cuts in the first three seconds make the answer feel costly.
• A rapid 0:09–0:25 filming montage turns the abstract need for scale into visible proof before the explanation.
• At 0:28 the double answers the mystery; the film then slows from 25 early shots to one 15.4-second closing shot for the recurring-series promise and CTA.

Hook formula: workload question + escalating answer: “How many {{REPEATED_TASKS}} today?” / “{{NUMBER}}, maybe {{HIGHER_NUMBER}}.”

Why it holds: There is no setup card. The dialogue starts at 0:00, the first cut lands at 1.29 seconds, and the camera goes to the hands exactly when the person counts. The format withholds how the work will get done.

Retention map: hook at 0:00 · filming-montage reset at 0:09 · mystery at 0:19 · double reveal at 0:28–0:29 · payoff at 0:32–0:42 · follow CTA around 0:58 · hard stop, with a next-day appointment rather than a visual loop.

Mix: Approximately 60% acted setup, montage and two-person conversation; 15% operator-and-double reveal; 25% locked direct address. All visible material is staged live action or a rendered avatar within photographed environments; there is no screen-recorded interface.

Rhythm: Seven cuts in the first 10 seconds; 25 total. The first 18 seconds are mostly 0.5–1.7-second inserts, then the conversation uses 1.5–4.9-second shots. The final 15.43 seconds are one held shot. Four overlays or frame changes occur inside shots; two changes near 0:08 are uncertain and should not be treated as proven cuts.

Rules this editor follows (inferred):

• Put a close detail on the counted answer, then return to a face for the consequence.
• Give each montage fragment a new room, outfit, or camera angle so the output feels numerous.
• Keep the double and original together in a wide shot before isolating the double.
• Slow the cutting once the mechanism is understood; let the spokesperson deliver the offer without interruption.

Full shot list

• Look: Bright, soft office and studio lighting, natural skin, white walls and plants; blue suit and saturated blue sculptural backdrop as the unifying element. Approximate palette from the sheets: deep blue #125384, teal-shadow #315C65, warm white #F3F1E8, charcoal #151A1F. The double swaps the suit for a fitted black shirt, making same-face/different-role readable.
• Captions: Closest free match Inter SemiBold or Poppins Medium. White #FFFFFF, sentence case, clean shadow rather than a box, no colored keyword treatment. Hook is larger, about 28–34 px at 720 px width; the remaining subtitles look about 18–24 px. Centered horizontally around 68–73% of frame height, typically 1–4 words per card; replace on speech rather than fly or bounce. Sizes are visual estimates.
• Graphics and overlays: At about 0:29, a small colorful abstract mark plus a white hand-drawn arrow and label point to the double. Keep this as the one explicit identity marker; recreate with an original icon and hand-drawn arrow, not the source logo.
• B-roll: Filmed vignettes of the same professional making different content: doorway, tripod, window, chair and elevator. Each lasts about 0.9–2.4 seconds until the mystery line gets 3.6 seconds.
• Transitions: Mostly hard cuts and shot/reverse-shot. The montage uses abrupt location and wardrobe changes; no evidence of a routine whoosh or elaborate wipe.
• Branding: Product identification appears inside the dialogue and arrow graphic; no separate full-screen end card in the sheets.

Measured: 177 words / 60.8 seconds = 176 wpm overall; roughly 245 wpm during voiced intervals. There are 31 pauses at least 0.25 seconds, median 0.5 seconds; the 1.56-second gap near 0:07.6 holds before the filming montage. The measured median pitch is 119 Hz with a 6.1-semitone standard deviation; 13 falling, 8 rising and 6 level sentence endings were classified. Questions rise; the reveal term around 0:28 and the CTA word around 0:58 receive extra emphasis.

Described: Conversational, quick and mildly comic in the exchange, then assured and inviting in direct address. The operator sounds overloaded and surprised; the double sounds calm, useful and a little playful. Exact accent is not essential to the format.

Voice direction: Use two clearly distinguished adult performances. The operator starts briskly, compresses the counting line, and leaves a beat before recognizing the solution. The second role answers with easy confidence, stresses the product's functional promise, and becomes more direct and upbeat for the final 15 seconds. Keep the lines short and natural, with a genuine question-and-answer rhythm. Use original voices; do not imitate the reference speaker.

• Music: A continuing tonal/music-like bed is measured between words at about −28.9 dBFS, only 7.3 dB below the measured voice level. Its exact instrumentation and tempo are not established by the sheets; choose a light, unobtrusive pulse (inferred) that lets dialogue carry the scene.
• Sound effects: Only 8% of cuts are louder than usual; there is no clear evidence of designed hits or whooshes on each cut. Use natural room/footstep texture if available.
• Silence and space: Leave the 0:07.6 pause before the first filming vignette and small reaction spaces at the 0:28 reveal and 0:42 handoff. The bed appears to continue through pauses.
• Mix targets: The reference measures −15.9 LUFS integrated and +0.7 dBFS true peak. For a new export, aim near −16 LUFS and keep true peak at or below −1 dBFS to avoid clipping.

• Title: Frame the outcome as a high-value professional getting a recurring content series without the recurring shoot. Template: “{{ROLE}} now shares {{VALUE}} every {{CADENCE}}—without {{OLD_BURDEN}}.”
• Cover / thumbnail: A mid-speech close-up of the blue-suited operator, mouth open, on a pale office background; a small white two-word subtitle across the lower torso. The face and blue jacket carry the image, not a headline card.
• Description, hashtags, pinned comment: The source post states the new daily series, credits the official avatar and links to an external story. No hashtag or pinned-comment pattern is available from the supplied evidence. For a new post, lead with the repeatable benefit and link to {{DESTINATION}}.
• Audience signals: 152,608 views, 494 likes, 500 comments and 327 reposts at same-day capture. No replay graph or comment text was supplied, so there is no evidence for which moment viewers rewatched or praised.

Target about 175 spoken words total. The budgets reflect the reference's 176 overall wpm and include space for questions and reactions.

[0:00–0:04 · workload hook · ~12 words]
COLLEAGUE: How many {{REPEATED_OUTPUTS}} today?
OPERATOR: {{NUMBER}}—maybe {{HIGHER_NUMBER}}. That's {{TIME_COST}}.

[0:04–0:08 · pressure · ~11 words]
COLLEAGUE: {{AUDIENCE}} wants more {{VALUABLE_EXPERTISE}}.
OPERATOR: All right. Let's start.

[0:09–0:19 · output montage · ~30 words]
Three clipped, original examples of {{TOPIC}}, each in a different room or outfit.
Keep each example independently useful but incomplete.

[0:19–0:25 · mystery · ~18 words]
OPERATOR: I had the whole afternoon blocked for this. Why do these look finished?

[0:25–0:32 · character reveal · ~21 words]
OPERATOR: Who are you—and how is this happening?
DOUBLE: I represent {{PRODUCT}}. I take on {{REPEATABLE_WORK}} while you {{HIGH_VALUE_WORK}}.

[0:32–0:45 · benefit, handoff · ~39 words]
OPERATOR asks one practical verification question.
DOUBLE answers with one concrete constraint or capability.
OPERATOR exits; DOUBLE turns toward the viewer.

[0:45–1:01 · recurring promise + CTA · ~44 words]
DOUBLE: When {{PROTAGONIST_ROLE}} cannot {{OLD_BURDEN}} every day,
{{PRODUCT}} still makes {{BENEFIT}} possible. Over {{DURATION}},
expect {{LANE_1}}, {{LANE_2}} and {{LANE_3}}.
Follow {{DESTINATION}} for {{CADENCE}}.

Script prompt (paste into an AI assistant with this blueprint):

Write a new 58–62 second, roughly 175-word dialogue-led launch film for {{PRODUCT}} about {{TOPIC}}. Cast an overloaded {{PROTAGONIST_ROLE}}, a colleague, and a visibly distinct second version or representative of the product. Use eight timed beats from this blueprint: workload question, pressure, three filmed samples, mystery, reveal, practical benefit check, handoff, and a stable direct-address promise. Make the product's actual capability and its limits accurate. Write short, speakable lines, six or fewer questions, no slogan-heavy narration, and one CTA to {{DESTINATION}}. Supply a shot and caption line for every beat. Do not reuse any source dialogue or identity.

Voice prompt (for actors or voice design):

Two adult conversational roles. The operator is quick, mildly overwhelmed and reactive; the product representative is calm, assured and lightly funny. Overall pace near 175 wpm with rapid short exchanges; pause before the reveal, then speak the final offer directly and warmly. Natural low-to-mid register, expressive questions, crisp consonants. Use original voices.

Visual prompts (one per recurring shot type):

• Office dialogue: Two adult colleagues in a bright modern workspace, medium close-up and reverse angle, soft window light, plants and neutral walls, one saturated {{ACCENT_COLOR}} jacket, candid conversation, 4:5 framing, empty lower-center area for captions.
• Work vignette: Same fictional professional filming {{TOPIC}} in a different office corner with phone or tripod visible, change jacket or room for each vignette, natural light, medium and wide alternation, no readable logos.
• Double reveal: Original professional and distinct {{PRODUCT_ROLE}} in one wide frame, contrasting wardrobes, sculptural {{ACCENT_COLOR}} background, believable eye contact, then a medium close-up of the second role; reserve space for a small handwritten identity arrow.
• Final address: Product representative centered in a locked medium 4:5 shot, blue/teal sculptural background, soft face light, natural hand gestures, direct eye contact, steady camera for 15 seconds.

Caption spec: Inter SemiBold or Poppins Medium; white #FFFFFF; 18–24 px at 720×900, with the first question 28–34 px; subtle shadow, no box; centered near 70% frame height; 1–4 words per chunk, timed to the words. Keep captions inside safe margins.

Music prompt: Light, modern, understated pulse or soft percussive ambient bed; stable through dialogue, no dramatic trailer swell, instrumental only; preserve the reaction pauses.

1. Lock the 60-second script and record separate operator, colleague and representative performances with clean dialogue.
2. Build the 0:00 question and counting answer first; cut to hands at about 1.3 seconds, then return to face at about 2.5 seconds.
3. Assemble 0:09–0:19 as short, varied filmed samples. Change setting or wardrobe on every cut and keep the spoken examples about the real {{TOPIC}}.
4. Place the first shared-frame appearance of the original and second role around 0:25; delay the arrow label until the identity line around 0:29.
5. Run the benefit exchange in shot/reverse-shot, then show the original leave while the representative remains.
6. Hold the final representative shot continuously from about 0:45 to the end; use delivery and hand movement for energy.
7. Add small white speech captions throughout, one identity arrow, a low bed and natural room texture. Balance dialogue, check sync, export 720×900 at 24 fps.

• Workload question starts at 0:00 and the time cost is understood by 0:04.
• At least three visibly different work examples arrive before the product is named.
• First shared frame of original and second role is clear; the identity label appears around 0:29.
• Dialogue states a specific {{BENEFIT}} without claiming capabilities {{PRODUCT}} does not have.
• About 25 cuts across 60 seconds; most early holds under 2 seconds, then a deliberate 12–16-second final address.
• Captions match speech, sit near 70% frame height, and do not hide hands or faces.
• Around 175 spoken words, with a real pause before the montage and reveal.
• Audio is around −16 LUFS; true peak stays at or below −1 dBFS.
• CTA names {{DESTINATION}} and a concrete posting or product cadence.

• The creator's signature (don't copy): A specific public figure, their likeness and voice, the source company's avatar, distinctive blue architectural set, logo, and real-estate content access.
• The transferable format: Begin on a recurring-work question; show multiple examples of the work; introduce a double or product representative as the answer; let the original hand off the burden; finish with a calm direct-to-camera recurring promise.
• Cast and record your own people, use authorized likenesses and claims, and source original or licensed footage and music.

Source: https://x.com/HeyGen/status/2108166496409248157 · measured with dissect 1.1.0 on 2026-10-08 · interpretations and visual estimates are identified in context.