Rapid visuals for a new-business pitch deck: campaign hero, product tiles, key-visual explorations, quote slides — all client-branded, all clearly-labeled-as-mockup, all iterable in under an hour before the pitch.

Brief + locked direction → full pitch-deck asset set. Speed first (pitch windows are brutal), reusability second (if you win, these become production inputs), clarity third (nobody should confuse a mockup for shipped work).

When to Use

Net-new business pitch, 48-hour turnaround
Mid-pitch concept swap — client leaned toward direction 3, need assets tonight
Multiple concepts per pitch (A/B/C) — each needs its own hero + 3 tiles
Internal review deck for a signed client's next campaign
Pre-approval creative exploration before any photo/video production spend

Do not use for finished deliverables — pitch mockups are intentionally aspirational and mockup-flagged. For final assets after win, use agency-multi-brand-pack or project-specific production workflows.

Prerequisites

Ask the user (one message):

Client slug + brand file — path to clients/<slug>/brand.md (output of agency-brand-scoping). If missing, run brand-scoping first.
Pitch concept(s) — one tagline or idea per concept (e.g. "Banking that waits for you")
Deliverable breakdown — how many heroes, tiles, quote slides per concept? Default: 1 hero + 3 tiles + 2 quote slides
Pitch date — drives whether we run cheap drafts or go straight to finals
Confidentiality — NDA status; affects Drive folder + any watermark choice
Mockup labeling — visible "MOCKUP" watermark, subtle footer, or none? (default: subtle footer for external pitches)

If the user hasn't locked a direction yet, stop and route them to agency-brand-scoping first. Don't pitch on guesses.

How to Run

1. LOAD BRAND    → read clients/<slug>/brand.md + brand-system.json
2. PLAN          → table of every asset with prompt + aspect + model
3. ESTIMATE      → gen-ai pricing on the full manifest (target < $5)
4. DRAFT PASS    → run at draft quality (gemini-3.1-flash-image / recraftv4) for fast iteration
5. CLIENT REVIEW → internal pitch-team review on the draft set
6. FINALIZE      → upgrade approved concepts to flagship models
7. WATERMARK     → add subtle "MOCKUP" footer where required
8. DELIVER       → organized folder + approved upload/link + deck-ready filenames

Rules:

Always draft-then-upgrade. 80% of pitch iterations die in internal review — don't pay flagship prices twice.
Every prompt cites the brand file. ` on every generate / batch.
Pitch mockups should be obviously mockups unless the client expects production-level polish. Padded aspect ratios, generic text ("HEADLINE HERE"), or visible mockup tags reduce the "why isn't this the final?" conversation.
Never use real competitor brand marks, real logos you don't have rights to, or recognizable faces. Stock-lookalike issues kill pitch trust.

Quick Reference

{
  "client_slug": "acme-fintech",
  "concept": "waits-for-you",
  "phase": "draft",
  "brand_system": "clients/acme-fintech/brand-system.json",
  "defaults": {
    "model": "recraftv4"
  },
  "variants": {
    "asset": ["hero", "tile-01", "tile-02", "tile-03", "quote-01", "quote-02"]
  },
  "mockup_footer": "MOCKUP — For internal review only"
}

Quick Reference

Sub-task	Draft	Final	Notes
Hero / key visual	`recraftv4`	`flux-2-pro`	Flux lifts it to photoreal polish when a concept sticks
Product / audience tiles	`recraftv4`	`recraftv4_pro` if available, else hold at `recraftv4`	Usually fine at draft quality
Quote slides (readable headlines)	`ideogram-v3`	`ideogram-v3`	Only model that renders type reliably
Animated key visual for video pitch	`kling-v3-standard`	`kling-v3-pro`	See `gen-ai-use` skill for assembly
Throwaway concept tests (10+ options)	`gemini-3.1-flash-image` / `gemini-3.1-flash-image`	n/a	~1 credit each

Procedure

Draft at 1/5 the cost of finals. Approve concepts on cheap drafts, upgrade 2-3 winners only.
One concept per folder. pitches/concept-a/, pitches/concept-b/ — never mix.
Watermark external pitches. Subtle MOCKUP footer on every image; use a deck/editor overlay, ImageMagick, or ffmpeg if available.
Reuse hero as tile input. gen-ai generate -m recraftv4 -i pitch/hero.webp --prompt "square crop variant" keeps concept coherent.
Never show the client 6 options with 3 winners. Curate to 2-3 strong concepts max before the pitch.
Keep filenames deck-ready. hero.webp, tile-01.webp — not gen_01H2XYZ.webp.
Lock the prompt file per concept. pitches/<concept>/prompts.json — when you win, production starts here.

Pitfalls

Over-polished mockups that the client expects you to deliver for pitch prices — add mockup footer + use draft models
**Skipping ` — palette drift mid-manifest, deck looks incoherent
Cross-client prompt contamination — don't copy prompts from another client's repo without stripping brand specifics
Using real faces / real brand marks — rights issues; use descriptors ("a founder in their 30s") and generic marks
Missing rights documentation — if the pitch wins, you'll need to reproduce these with clean provenance; save results.json per run
No gen-ai pricing before a 50-variant concept explosion — pitch budgets don't absorb $40 surprises

Verification

Run gen-ai whoami to confirm authentication, then re-run the failed command with --debug.

Step 1: Plan the asset set

Present a table before spending:

#	Asset	Concept	Model (draft → final)	Aspect	Notes
1	Hero	"Banking that waits"	`recraftv4` → `flux-2-pro`	16:9	Full-bleed, tagline overlay
2	Tile A	Product focus	`recraftv4`	1:1	Clean product shot
3	Tile B	Audience focus	`recraftv4`	1:1	Founder at desk, moody
4	Tile C	Benefit focus	`recraftv4`	1:1	Abstract benefit icon
5	Quote slide 1	Customer voice	`ideogram-v3`	16:9	Readable headline type
6	Quote slide 2	Founder voice	`ideogram-v3`	16:9	Readable headline type

Estimate the manifest. If under $5, proceed. If over, cut the draft resolution or concept count.

Step 2: Draft manifest + run

CLIENT="acme-fintech"
CONCEPT="waits-for-you"
OUT="clients/$CLIENT/pitches/$CONCEPT/drafts"
mkdir -p "$OUT"

cat > /tmp/pitch-$CLIENT-$CONCEPT.json <<EOF
{
  "defaults": {
    "model": "recraftv4"
  },
  "jobs": [
    { "id": "hero",       "prompt": "$CLIENT hero, concept: banking that waits for you — editorial, cinematic, headline overlay area left-third", "aspectRatio": "16:9" },
    { "id": "tile-01",    "prompt": "$CLIENT product tile — clean product render, brand palette",           "aspectRatio": "1:1" },
    { "id": "tile-02",    "prompt": "$CLIENT audience tile — founder at desk, moody lighting, documentary", "aspectRatio": "1:1" },
    { "id": "tile-03",    "prompt": "$CLIENT benefit tile — abstract waiting metaphor, minimal composition", "aspectRatio": "1:1" },
    { "id": "quote-01",   "prompt": "quote slide, large editorial headline 'Finally a bank that moves like I do.', brand palette, minimal background", "aspectRatio": "16:9", "model": "ideogram-v3" },
    { "id": "quote-02",   "prompt": "quote slide, large editorial headline 'Built for founders who don't wait.', brand palette, minimal background",   "aspectRatio": "16:9", "model": "ideogram-v3" }
  ]
}
EOF

gen-ai batch run /tmp/pitch-$CLIENT-$CONCEPT.json -c 4 -o "$OUT"

Review internally. For approved concepts, re-run the subset with model: flux-2-pro and output to pitches/$CONCEPT/finals/.

Cost & time

Phase	Spend	Time
Draft pass (6 assets, `recraftv4`)	~$1-2	10-15 min
Final upgrade (2-3 winners to `flux-2-pro`)	~$2-4	15-20 min
Total pitch kit	~$5	~45 min end-to-end

Compare to production-phase cost after win: single campaign set with final photography-grade generations runs $20-100+.

agency-pitch-mockups