Jordan Cameron — AI Asset Hub

0 · The results — start here

Everything below this block is an ingredient. This is what the ingredients make. Every clip here was generated from the assets on this page — no camera, no editor, no studio.

How they're made: voice → section 1 · identity images → section 2 · video routes → sections 4–5 · the explainer method → section 6 · plan your own shots → the shot planner.

1 · Voice — Jordan's clone

Generate speech in Jordan's voice — copy, paste, run

curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/Wrh70uw8jFy1g5IViE35" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "YOUR SCRIPT HERE",
    "model_id": "eleven_v3",
    "voice_settings": { "stability": 0.0, "use_speaker_boost": true }
  }' --output speech.mp3

Rule of thumb: eleven_v3 for anything client-facing (keeps the accent, expressive). Never eleven_multilingual_v2 for Jordan — it americanizes him.

2 · Images — AI Jordan, ready to use

Generated with gemini-3-pro-image (Nano Banana Pro) at 2K, identity-locked to three reference frames from the welcome video. Right-click → save, or copy any of these into ChatGPT/Gemini as the reference for new variations.

Studio headshot · 1:1 · LinkedIn / avatar

Natural headshot · 1:1 · the trustworthy-adviser look, light-grey seamless

Office editorial · 16:9 · website / about

Thumbnail: arms folded · 16:9 · navy backdrop, right two-thirds clean for text

Thumbnail: house key · 16:9 · left third clean for text

Premium portrait · 3:4 · print / cover

The reference frames (ground truth identity — attach these when generating new images)

ref @ 4s

ref @ 16s

ref @ 28s

Need a fast draft instead of final quality? gemini-3-1-flash-lite-image (Nano Banana 2 Lite) returns in ~4s — good for iterating on composition, not for the final render.

3 · Thumbnail factory — copy-paste prompts

Paste a prompt into ChatGPT (image mode) or Gemini, attach 1–2 reference photos from section 2, fill the {{PLACEHOLDERS}}, hit go. Nano Banana Pro (gemini-3-pro-image) gave the best facial likeness in our testing. The expression rule is already written into each prompt.

Using the attached photos as the exact likeness reference (same man, same face), create a YouTube thumbnail, 16:9, 2K: he stands waist-up on the LEFT third, arms loosely folded, calm assured closed-mouth smile, looking straight at camera — composed, never exaggerated. Right two-thirds a clean deep navy (#003F5E) studio backdrop where the headline "{{TEXT ON THUMBNAIL}}" appears in huge white sans-serif. Crisp key light, premium and calm, high contrast but tasteful.

Using the attached photos as the exact likeness reference, create a 16:9 social banner: he sits at a modern desk reviewing mortgage documents, warm natural light, subtle relaxed smile with mouth closed. Overlay the title "{{VIDEO TOPIC}}" in clean bold type on the upper right with plenty of breathing room. Total Mortgages brand feel: electric blue (#1421FF), deep navy, white, warm neutrals. Photorealistic, editorial quality, quiet confidence.

Using the attached photos as the exact likeness reference, create a 16:9 thumbnail: he holds a single house key at chest height with a small warm closed-mouth smile, blurred sunny New Zealand home exterior behind him, left third kept clean where the text "{{TEXT ON THUMBNAIL}}" appears in massive bold white-on-navy type. Professional finance-brand grade, natural colors, composed energy — not salesy.

4 · Video — the six production routes

Different jobs need different models. Prices and model IDs are exact — this table is the menu.

5 · The B-roll engine — Gemini Omni Flash, explained

This is the newest capability on the page and the biggest unlock for Total. gemini-omni-flash-preview is Google's new any-to-video model (announced alongside Nano Banana 2 Lite). It runs on Google's new Interactions API — the classic API rejects it — and it accepts almost any combination of inputs:

Proven by demo — nine shots · 76 seconds · $7.60 all-in

Plus the couch-to-window walk in section 4. Dialogue and snap are second-round takes: the first versions read too animated for Jordan, so the prompts now carry an explicit reserved-delivery block ("does NOT grin, minimal hand movement, not animated") — same rule as the section-2 images, and it works just as well in video.

The identity boundary — found, then fixed, same day

Run it yourself — Google's official skill scripts

# one-time: clone google-gemini/gemini-skills → skills/gemini-omni-flash-api
# needs: pip install "google-genai>=2.10.0" && export GOOGLE_API_KEY=...
python video/generate_video.py \
  "The man from the reference image rises from the white couch and walks slowly \
to the large window, camera follows in a smooth arc from front to profile. \
Soft natural daylight, calm ambient room tone only, no speech, no music." \
  --image images/refs/jordan_ref_4s.jpg \
  --aspect-ratio 16:9 --duration 8 \
  --output broll_couch_to_window.mp4
# model: gemini-omni-flash-preview · Interactions API · $0.10 per second of video

Also available via fal (fal-ai/gemini-omni-flash) — we buy it from Google direct: faster in our testing and one fewer intermediary. Prompt discipline: always say "ambient room tone only, no speech, no music" unless you want the model inventing a soundtrack.

6 · Explainer route — Vox-style motion graphics

Not every video needs Jordan's face on camera. For educational content — "What the OCR cut means for your fixed rate", "First-home buyer, step by step" — the highest-trust format is the Vox-style animated explainer. Explainers №1 and №2 are done:

The method behind it (proven by motion designers working entirely in Claude Code + Remotion) — reproducible for any topic with the assets already on this page:

1. Script = the timeline. Write the voice-over first, as a table: each VO beat gets a row with its foreground asset, midground asset, and the image prompt that creates them. The narration drives everything.
2. Lock the visual system. One shared background across every scene, one font, one accent palette — for Total that's already decided: navy #003F5E, electric blue #1421FF, white. A locked background makes cuts feel like one continuous shot.
3. Three layers per scene. Static background / midground cutouts (people, landmarks — converted to a print-style halftone treatment by prompting Nano Banana: "make this black and white with a halftone pattern") / foreground props (houses, charts, numbers). One folder per scene.
4. Animate in plain English. No After Effects, no keyframes: tell Claude Code "the house springs up first, then Jordan's cutout, staggered — give each cutout an offset electric-blue marker stroke behind it." It translates to Remotion's spring() and interpolate().
5. Fine-tune with prop controls. Ask for a Remotion Studio prop control on every element — then drag scale/X/Y live and save the numbers. No re-rendering to reposition.
6. Assemble to the voice. Generate the VO with Jordan's clone (section 1), then: "embed the voice-over and sequence the scenes to it — each scene starts and ends on its own narration line."
7. Polish and render. Music + light foley last, then "render the whole thing as 1080p MP4." Scrubbing audio sounds jerky in Studio; the render is clean.

7 · Price sheet — what a video actually costs now

Exact models, exact unit prices, and what a typical asset comes to. This is the whole menu on one card.

Read it like this: a month of content — 4 personalized client messages, 2 explainers, an intro B-roll pack of 6 shots — is about $15 of compute. The scarce input is now the script and the taste, not the production.

8 · The pipeline — how the assets combine

9 · Next assets to add