Brief your offer once, lock your look and AI creator cast, then batch 10+ ad variants from a single storyboard pass — no agency, no shoot day.
This guide is reviewed for clarity, service accuracy, and AI-search readability. The next quarterly content review is tracked internally before unsupported metrics or client proof are added.
Why Most Founders Use AI Video Wrong
Most founders treat AI video tools like a vending machine. One prompt, one clip. Repeat every time.
That is faster manual labor, not a workflow. Every ad means re-typing your offer, re-describing your visual look, and re-casting your presenter from scratch. You are doing the agency's job without the agency's output.
invideo Agent One works differently. It has a persistent Context — a project memory that holds your offer brief, visual style, and creator roster across every session. Brief it once. Every ad it generates afterward inherits that foundation automatically.
This playbook walks you through all five stages. Run it once and your ad operation runs like a pipeline.
Tool Setup: invideo Agent One Workspace
Get your workspace ready before Stage 1.
- Go to invideo.io and create an account. Paid plans start at $25/month (Plus, 50 AI minutes) or $60/month (Max, 200 AI minutes). The Max plan is the right level for running Stage 5 batches regularly. A fourth Generative tier ($120/month) adds access to premium video models if you need higher-fidelity output.
- Inside Agent One, locate the Context panel — this is where all persistent memory lives. It auto-loads at the start of every session in that workspace.
- Context is stored per workspace. Create a separate workspace for each product or brand line so briefs do not bleed across offers.
- To update Context without re-briefing from scratch: open the Context panel, edit only the field that changed (offer price, CTA, visual style), and save. The agent inherits the update on the next generation.
- To reset Context: clear individual fields rather than deleting the workspace. This preserves your locked items in every other field.
- Agent One supports vertical (9:16), square (1:1), and horizontal (16:9) output. You lock your format mix in Stage 5.
Quick-Start Checklist: Lock These Five Items Before Stage 3
If any of these five Context items are missing when you reach Stage 3, your creator outputs will be generic and inconsistent. Check all five before you continue.
promptPRE-STAGE 3 CONTEXT CHECKLIST [ ] Offer brief saved to Context (Stage 1 output) [ ] Target audience defined in Context — one specific person, not a demographic [ ] Brand voice locked in Context — 3 adjectives max (e.g., direct, warm, credible) [ ] Winning visual style saved to Context (Stage 2 output — one locked look only) [ ] Hook angle confirmed — the emotional or curiosity-led entry point your audience responds to
Do not skip this check. The agent inherits whatever is in Context. Empty fields produce generic output every time.
Stage 1: Lock Your Offer Brief Into Context
This is the one-time setup that makes everything else work. You are not writing an ad here. You are writing the operating instructions for every ad this agent will ever make for this offer.
Paste this into invideo Agent One's Context input:
promptOFFER BRIEF — PERSISTENT CONTEXT Offer: [Product or service name + core transformation in one sentence] Price point: [$X one-time / $X per month / free trial available] Audience: [One specific person — example: "a solo founder running paid social ads without a team, spending $500–2,000/month, getting inconsistent results"] Core problem solved: [The single biggest pain this product removes] Voice: [3 adjectives only — example: direct, credible, no-fluff] Hook angle: [Emotional or curiosity-led entry point — example: "Most founders waste 80% of their ad budget on creative that was never tested"] CTA: [The one action you want the viewer to take — example: "Start free at [URL]" or "DM the word DEMO"] What makes it different: [One sentence, no superlatives — example: "The only tool that holds your offer in memory so you never re-brief it"] What it is NOT: [One sentence ruling out misfit buyers — example: "Not for agencies managing 10+ clients"]
Save this to Context. You will not type your offer again. Every slide, every scene, every line the agent writes from here pulls from this brief automatically.
Stage 2: Generate a Dozen Visual Styles, Lock the Winner
You need a consistent look before you build a creator cast. This stage gives you twelve options in one pass so you make a deliberate choice rather than guessing.
Paste this into Agent One as a generation prompt — not a Context update:
promptVISUAL STYLE GENERATION PROMPT Using the offer brief in Context, generate 12 distinct visual ad styles for this product. For each style, output: - Style name (2–3 words) - Color palette (primary, secondary, accent) - Typography mood (example: bold sans-serif, editorial serif, handwritten) - Scene mood (example: studio clean, lifestyle b-roll, documentary raw, bold graphic) - Lighting direction (example: high-key bright, moody low-key, natural window) - One example scene description for the Hook beat Vary across: clean studio, lifestyle b-roll, UGC-style, bold graphic, documentary, animation. Do not repeat a style. Label them Style 01 through Style 12.
Decision rubric — pick one style by answering these four questions:
- Does it match how your audience already consumes content on this platform?
- Is it reproducible without a shoot day? Avoid styles that require live footage you do not have.
- Does it make the offer feel credible, not just visually interesting?
- Can you describe it in two sentences so the agent regenerates it consistently?
Once you pick a style, add it to Context:
promptVISUAL STYLE — ADD TO CONTEXT Locked style: [Style name from the list above] Color palette: [Primary / Secondary / Accent — hex codes or clear descriptions] Typography: [One sentence description] Scene mood: [One sentence description] Lighting: [One sentence description] Avoid: [One sentence on looks that break the brand feel]
Stage 3: Build Your AI Creator Roster and Lock It
AI creators — on-brand presenter personas or avatars — need to be defined once and stored in Context. Without this step, the agent casts whoever fits the prompt. Your ads look like they come from different brands.
promptAI CREATOR ROSTER PROMPT Using the offer brief and visual style in Context, define a roster of 2–3 AI creators for this ad system. For each creator, output: - Creator code name (example: ALEX, SAM, MORGAN) - Gender and approximate age range - Appearance: skin tone, hair, build — 2–3 sentences, specific enough to regenerate consistently across sessions - Wardrobe: what they wear in ads for this offer — be specific (example: "plain fitted white tee, no logos, light wash jeans") - Delivery style: how they speak and move on camera (example: "direct eye contact, minimal body movement, speaks like a peer not a presenter") - Best use: which hook type or ad format this creator fits best (example: "ALEX for problem-first hooks; MORGAN for transformation reveal hooks") These creators must feel like they belong to the same brand. Do not make them look like generic stock footage talent.
After the agent returns the roster, save the full output to Context under a "Creator Roster" field. From this point forward, every ad you generate references these creators by code name. No re-casting.
Stage 4: The Beat-by-Beat Storyboard Template
Every ad follows four beats: Hook, Problem, Product, CTA. This template auto-inherits your locked look and creator. You fill in content — not format decisions.
promptBEAT-BY-BEAT AD STORYBOARD OFFER: [Auto-inherited from Context — confirm: your offer name] CREATOR: [Code name from locked roster — example: ALEX] FORMAT: [9:16 vertical / 1:1 square / 16:9 horizontal] PLATFORM: [Instagram Reels / TikTok / Meta Feed / YouTube Pre-roll] --- BEAT 1 — HOOK (0–3 seconds) Hook type: [Pattern interrupt / Bold claim / Question / Relatable pain] Opening line: [Exact first sentence the creator says or text on screen] Visual: [What is happening on screen — use the locked visual style language] --- BEAT 2 — PROBLEM (3–8 seconds) Core pain: [The specific frustration your audience has right now] Delivery: [How the creator says it — empathetic / blunt / matter-of-fact] Visual: [Cut, graphic, or b-roll that reinforces the pain — stay in locked style] --- BEAT 3 — PRODUCT (8–18 seconds) What it does: [One sentence — lead with the outcome, not the feature] Proof element: [What makes this credible — demo clip, result stat, before/after] Delivery: [Tone shift here — more certain, more energized than Beat 2] Visual: [Screen recording, product shot, or graphic — describe specifically] --- BEAT 4 — CTA (18–30 seconds) Action: [The single thing you want the viewer to do] Line: [Exact words the creator says or text overlay] Visual: [Static end card / creator direct-to-camera / animated text] Urgency: [Optional — only include if it is real, not manufactured scarcity] --- STORYBOARD NOTES FOR AGENT ONE: Generate this ad using the locked offer brief, visual style, and creator roster in Context. Output a full scene-by-scene video script with shot descriptions ready for AI generation.
Feed the completed storyboard to Agent One as your generation prompt. It produces a full scene-by-scene script aligned to everything locked in Context.
Stage 5: Batch 10+ Variants From One Locked Ad
Once you have a complete storyboard from Stage 4, you do not write another brief. You spin it into ten variations in a single pass.
promptBATCH VARIATION PROMPT Using the storyboard in this session and the locked offer brief, visual style, and creator roster in Context, generate 10 ad variations from this single storyboard. Vary the following across all 10 outputs: - Hook type: rotate through pattern interrupt, bold claim, question, relatable pain - Format: include at least 3 vertical (9:16), 2 square (1:1), and 1 horizontal (16:9) - Hook line: rewrite the opening line for each — same offer, different entry angle - Creator: rotate between locked roster members where the hook energy calls for it Do NOT change: - The core offer or CTA - The locked visual style - The product beat (keep the same proof element unless told otherwise) Output: 10 numbered variation briefs, each with a new hook line, format label, creator code name, and any beat-level changes from the original. Label them V01 through V10.
One storyboard becomes ten testable ads. Feed whichever variations perform back into Context as a "proven hook" reference so future storyboards start from what your audience already responded to.
What a Full 5-Stage Run Actually Produces
Here is what to expect from a complete run.
Time investment:
- Stages 1–3 (first-time setup): 45–90 minutes total. Every subsequent product uses the same structure and takes under 30 minutes.
- Stage 4 (one storyboard): 10–20 minutes to fill in the template and generate a complete scene script.
- Stage 5 (batch): One prompt returns 10 variation briefs in a single generation pass.
Output per run:
- 10 production-ready variation briefs covering at least three formats (vertical, square, horizontal).
- Each brief is immediately renderable inside Agent One — no reformatting, no re-briefing.
- After the first run, Stages 4 and 5 are the only repeating work. The setup in Context is permanent until you deliberately change it.
Plan math (based on invideo's published figures):
- The Plus plan (50 AI minutes/month at $25/month) covers approximately 35–40 finished short-form clips per month.
- The Max plan (200 AI minutes/month at $60/month) gives you roughly four times that capacity — more than enough to run multiple Stage 5 batches per month.
What disappears from your workflow permanently:
- Creative director briefs
- Video shoot scheduling
- Re-explaining your offer to a freelancer
- Re-casting presenters on every ad
At Hypd, this is the workflow we point founders to when they are ready to stop one-shotting prompts and start running an actual ad system.
FAQ
Does invideo Agent One require a paid plan?
Yes. The free tier adds a watermark to every export and limits weekly output volume — neither works for ad production at scale. The Plus plan ($25/month, ~35–40 clips/month) is sufficient for testing this system. The Max plan ($60/month, ~140–160 clips/month based on invideo's published AI minute figures) is the right level once you are running Stage 5 batches regularly.
What happens if my offer changes — do I have to redo all five stages?
No. Open the Context panel and edit only the fields that changed — a new price point, a revised CTA, a new hook angle. Save and continue. Every field you did not touch stays locked. If your visual style or creator roster is still accurate, those carry forward automatically. You only repeat the stages where the inputs actually changed.
How many ad variations can I realistically run before hitting plan limits?
On the Max plan (200 AI minutes/month), you have room for multiple full Stage 5 batches per month based on invideo's published per-clip estimates. Track your AI minute balance in the invideo dashboard after each batch run so you know your remaining capacity. If you hit the ceiling mid-month, the Max plan's 200-minute allocation resets on your billing date — or you can upgrade to the Generative tier ($120/month) for a higher ceiling plus access to premium video generation models.