AI-Native Creative
Sora, Runway, Veo, Flux, Midjourney, ElevenLabs, HeyGen and friends, used to 10x creative velocity.
- Version
- v2026.10.1
- Updated
- Oct 24, 2026
- Size
- 11 KB
What you get
Which AI creative tools to use for what, the economics against real production, platform policies, when human UGC still wins, and a production workflow for a scaled paid account. Pairs with the Creative Production System for the testing matrix and Voice of Customer for the scripts.
Skill content
DemandBox — AI-Native Creative
The creative economics of paid media changed in 2024–2025. A theme that used to take 2 weeks to produce now ships in 48 hours. The accounts that will win the next 24 months are the ones that built the AI production muscle before competitors did. This skill is that muscle.
First principles
- AI is a velocity multiplier, not a quality shortcut. It lets you test 10× more ideas. The winning ideas still come from humans (and
voice-of-customer+creative-production-system). - Ship when the AI is good enough, not when it's perfect. A "good enough" AI ad that runs tomorrow beats a perfect human ad that runs in six weeks. Paid media is about iteration, not craft.
- Disclose when required, don't disclose when not. Meta requires disclosure for AI-generated/altered content depicting real events or real people. B2B ads with AI-generated stock imagery usually don't require disclosure — but know the platform policy before shipping. See platform notes below.
- Know when NOT to use AI. Human UGC still outperforms AI UGC for high-trust categories. Real customer testimonials beat synthetic every time. Founder-to-founder doesn't work AI-generated — it's meant to feel human.
Tool stack (as of 2026)
Choose by job-to-be-done, not by hype.
Video generation
| Tool | Best for | Cost profile | Limitations |
|---|---|---|---|
| Sora (OpenAI) | Cinematic short-form, B-roll, concept pieces. Strong for atmospheric DTC. | API + per-second pricing | Availability still gated; usage policies strict on public figures |
| Veo 3 (Google DeepMind) | High-quality 8-second clips with native audio. Best current quality for realistic video. | Gemini subscription or Vertex API | Output length limits; complex scene control is still iterative |
| Runway Gen-4 | Motion-heavy B-roll, stylized effects, image-to-video. Best editorial workflow (character locking, scene-to-scene consistency). | Per-credit / subscription | Slight uncanny valley on faces; avoid close-up hero shots of people |
| Pika | Fast, cheap, disposable variants — ideal for TOFU test volume | Monthly subscription | Lower quality ceiling than Sora/Veo |
| Kling | Strong on realistic human motion | Credits | Chinese-origin tool — check client policy before production use |
| Luma Dream Machine | Quick ideation and motion concepts | Subscription | Best as a sketchbook, not final |
Static image generation
| Tool | Best for | Notes |
|---|---|---|
| Midjourney v7 | Stylized hero visuals, editorial aesthetic | Still the quality leader for art direction |
| Flux 1.1 Pro / Flux Kontext | Photorealism, product shots, in-context edits | Ship-ready for most static ad needs; excellent hands and text |
| DALL-E 3 / gpt-image-1 | Fast concepting inside ChatGPT workflow | Good for ideation, not finals |
| Ideogram 3 | Typography-in-image (ad headlines as visual) | Unique capability; useful for static ads with integrated copy |
Voice and audio
| Tool | Best for | Notes |
|---|
Sign up free to read the whole skill and download it
We'll email you a sign-in link and bring you right back here.
Version history
v2026.10.1 · Oct 24, 2026
First public release.