Creative Space

Nội dung

Provide: Product images, approved product facts, target platform/ratio; for unboxing also packaging photos, accessory list and actual opening order.

Receive: A product video (default 15–30 seconds, 9:16 or 16:9), storyboard and fact card.

Formats

  • Scene-based ad: Crosscut identifiable product-in-use shots with function/atmosphere shots; finish on product or packaging. Suitable for brand feeling and function.
  • POV unboxing: Hands and product only; first reveal → details/first use → closing, with tactile sounds. Suitable for discovery and physical appeal.

Tools and references

Use host reading tools for supplied product-page claims, specifications, price, product or packaging. Use this document's marketing beats and shot requirements. Flova generates settings/props, keyframes, video, music and narration; FFmpeg assembles/mixes/exports. Both are required.

  • Captions (optional): If requested, offer a suitable HyperFrames, Remotion or FFmpeg post-production workflow.

Read in steps 3–5; for that format.

Boundaries

Product images define SKU/packaging version, proportions, color/material, structure (buttons, ports, zippers), logo/text placement/content and condition. Pass actual photos as references; never invent a similar substitute, and request missing views. Use only supplied claims/numbers and real first-use actions, without invented functions, outcomes or reviews.

Opening order, packaging and accessory counts must match photos/lists. Without packaging photos, explain that an ordinary shipping box will be used instead of inventing branded packaging. Keep existing product and packaging labels faithful to the source photos. Add benefit, price or logo overlays only when requested, using verified facts. Use authorized images, logos and music; remind the user of applicable AI labels. Check current platform duration, ratio and safe areas before production.

Workflow

Channel/product (approval) → storyboard (approval) → element images (approval) → one test shot (approval) → batch shots (approval) → sound → assembly → inspection/delivery.

1. Confirm brief — approval

Collect product images/facts, benefits, platform, ratio and duration. Read supplied product pages with available browser/connectors. For unboxing collect packaging, accessories and opening sequence. Save the factual constraints above in docs/brief.md and confirm them.

2. storyboard — approval

Use this document's storyboard guidance marketing structure; label beats hook, turn, proof, cta and their contribution.

For scene ads, elements include the actual product, at least two settings that demonstrate advantages, and key props; each setting is separately reusable. Crosscut product-in-setting proof (identifiable SKU and use) with function/atmosphere hooks/transitions, which may omit the product if explicitly planned. Start with atmosphere/result, prove benefits, then finish on product or packaging. One generation task may contain several internal cuts: roughly 3–6 for 15 seconds or 6–12 for 30. Favor hard cuts, short beats and varied framing/movement.

For unboxing use the reference's three acts/touchpoints; every benefit needs visible evidence. Choose immersive-slow or brisk pacing. Per shot record duration, framing, action, reference images and sound. Obtain approval.

3. Element images — approval

Scene ads only: generate setting/prop sheets; use supplied photos for the product itself. Follow the generation/checks reference. Unboxing skips this step and directly references product/packaging photos.

4. One test shot — approval

Choose mode/model through Flova Skill, usually Quick because FFmpeg assembles. Test the hardest product reveal, first-use action or complex scene; approve product fidelity, lighting, action and sound before batching.

5. Batch — approval

Pass every reference used by each shot. Exclude music, subtitles and onscreen text. Check product fidelity, visible benefits and storyboard rhythm. Redo only failures, retaining approved work; if two corrections still fail, explain and agree on simplification or a model change.

6. Sound

Use at least one music track across the full film. If a scene ad has narration, write it before fixing duration. Align unboxing effects to physical actions; usable generated effects may be retained/reintroduced, otherwise mute and replace them in post. Let the user listen.

7. Assembly

Mute generated video tracks, assemble with FFmpeg, then add music, narration and selected effects; duck music under speech. For multiple ratios, generate and lay out each separately rather than simply cropping the finished ad.

8. Inspection and delivery

Watch end to end. Check product appearance, packaging, accessories and existing labels against the fact card; check synchronization and the ending. Deliver final video, storyboard, docs/brief.md, draft post copy with an AI-label reminder, and the actual project files. Redo only affected parts for revisions.

video-<product>/
├── source/    product/packaging photos, accessories, page facts
├── docs/      brief.md, storyboard, post copy
├── elements/  settings/props for scene ads
├── shots/     footage
├── audio/     music, narration, effects

└── output/    final video and assembly script

Generation requests and inspection

Use Flova Skill's general structure, with these product-specific rules taking precedence; actual model capabilities govern parameters. For Chinese-speaking users write prompts in Chinese; narration/dialogue follow the output language. Retain English style anchors, model-specific phrases and supported wrappers.

Images

Brief a director/colorist rather than merely listing contents: natural sentences for settings, short phrases for style/color/light/composition; no literary or invisible details. Integrate cinematic treatment into one flowing description: a specific creator/work as style anchor; composition/framing; key light/contrast; restrained grading (about 90% one dominant tone, no red/blue neon unless requested); overall texture/detail; and an expressive instant rather than a posed scene. Exclude faces; use hands, arms or backs when people are needed.

Keep existing product and packaging text faithful to the references. Add product-name, price or benefit overlays only when requested. Element sheets lock shape, material, color, dimensions, condition and distinctive marks; user photos define the product. Settings specify mood and the supported advantage. Keep hands, nails and cuffs consistent.

For keyframes, locate the storyboard instant and describe only its static state: pre-action pose/environment/camera for a start frame; completed action and next-shot preparation for an end frame; strongest standalone visual instant for a highlight frame.

Video requests

  1. Identify every product/setting/prop reference and its role, using the actual tool's reference syntax.
  2. Describe camera (push/pull/orbit/cut) → subject action/beat → space/position/depth → audio (dialogue/effects). Specify body part, amplitude, speed and force.
  3. If the tool supports long multi-cut requests, mark “shot 1,” “shot 2” chronologically; avoid relying on precise per-cut seconds the model cannot reliably execute.
  4. Use music/effect/dialogue wrappers only when supported by the selected model; otherwise identify them in ordinary text.
  5. Omit narration text when narration has a separate track; exclude music in nearly every shot, and subtitles/onscreen text in every shot.

Inspect each stage

  • Elements: Product appearance/structure/color/material/logo matches photos. Each setting fits the storyboard and supports at least one advantage; settings collectively cover at least two advantages. Style and any visible text are correct.
  • Keyframes, when used: Correct storyboard moment and inherited product, setting and props, without deformation; check space for text only when text packaging has been agreed.
  • Video: Identifiable SKU/use, expressed function/atmosphere, planned actions/camera, energetic cuts, product continuity, no unwanted music/narration/subtitles, correct product or packaging treatment, supported duration.
  • Audio: At least one music track spans the film; voice, tone, wording and language match approval.
  • Final: Intercut the two shot types with hard cuts and visibly different compositions; preserve product and packaging accuracy and the agreed ending. Check any requested text/logo overlays for content, placement and duration.

For blocking, entitlement or availability failures, use another video model with equivalent multi-reference support and the same references; tell the user which changed.

POV unboxing

Show hands and product, never faces.

Three acts

ActShare15 s30 sPurpose
First reveal~20%3 s6 sPackaging → opening → first product reveal, building anticipation
Details/first use~60%9 s18 sAppearance, accessories and first operation, building trust
Closing~20%3 s6 sArranged product, one conclusion and next action

Start with action immediately: hand already on packaging or blade opening tape, rather than an empty establishing frame.

Real physical touchpoints

Choose close-up actions supported by product/packaging photos: film peeling, tape cutting, lifting lids, sliding an inner drawer, untying ribbon, magnetic closure; pressing buttons, unscrewing caps, unzipping, clicking clasps, turning controls or pages; removing accessories individually from inserts and arranging them. If the opening mechanism is unclear, ask rather than invent it.

Camera, hands and lighting

Simulate the viewer's eyes, slightly downward, with hands entering from below; only hands/product/surface appear. Specify one/two hands, action and grip (edge pinch, palm support, fingertip movement). Keep skin, nails and cuffs consistent.

Use natural slight handheld motion. For close-ups, gently rotate the product to move reflections. Avoid slider pushes, aerials and gimbal smoothness. Move from tabletop wide → product close-up (60–80% of frame) → use close-up (hands/product/action) → material/craft macro; adjacent shot sizes should not jump more than two levels.

Opening shots usually stay below three seconds; immersive slow unboxing can use 4–5, first use 5–6 to preserve complete actions. Prefer slightly warm window light around 45° from the side, soft natural shadows and a real desk/bed/floor with light lived-in detail. Avoid visible studio-light reflections, seamless black/white backgrounds and perfectly uniform lighting; preserve natural, unblown highlights.

Details and first use

Include at least one appropriate detail and first-use shot:

CategoryDetailsFirst use
ElectronicsPorts/buttons/bezel/materialPower on, pair, first operation; clear screen contents
Beauty/skincareBottle, texture, cut surfaceDispense, take product, spread on back of hand
Clothing/shoes/bagsFabric, stitching, hardware, liningLift/unfold, touch fabric, operate zipper
Food/drinkSeal, contents/granulesPour, brew, lift one piece
Home/kitchenMaterial, edges, storage structureAssemble, place on counter for first use
Toys/collectiblesPaint, joints, accessoriesUnpack, pose, activate light/sound

Sound and prompting

Align tactile ASMR precisely: film, cardboard friction, tape, magnetic closure, zipper, button and cap. A video model may generate them together; replace mismatched sounds in post. An immersive version may have no speech; optional narration is quiet, conversational and shorter than the visual action. Keep music below narration and tactile effects.

Write requests in this order:

  1. Reference roles: first image defines unchanged product shape/color/material/logo; second defines actual packaging and opening mechanism. Describe hands/surface in text.
  2. View/environment: first-person, slightly downward, e.g. wooden desk beside a window with natural light from the left.
  3. Start → action → end, e.g. hold box sides → slowly push lid open with thumbs → reveal product and hold for half a second.
  4. Framing, handheld motion and focus.
  5. Which sounds occur when; explicitly no speech if there is no narration.
  6. Preserve appearance; exclude faces, text, captions, watermarks and items absent from the real package.

Storyboard and asset handoff

When new shot planning is needed, the host uses this workflow directly; reuse a clear existing shot list or edit brief. Record each S01 segment’s script link, visual/information goal, duration/time, subject/action, camera or element motion, exact text/sound, transition and production method. Link E01 target elements to user material and actual resources; keep REF-E01 source identities separate and mark unavailable assets as pending. Check identity, space, action and sound continuity; retime to actual speech. For marketing, label applicable hook/turn/proof/cta roles and the viewer gain, using grounded demonstrations; teaching and narrative work retain their own structure. For an unclear expression, find a real reference and record what to borrow/replace. No separately named planning skill or file is needed.

Quy trình sáng tạo

  1. Flova

    Flova

    Generate product shots

    Create product close-ups, use scenarios, and unboxing visuals.

  2. FFmpeg

    FFmpeg

    Assemble the video

    Edit the shots, add audio, and export the video.