We Tested Google Veo and Runway to Create This AI Film. It Was Wild. | WSJ

🎬 What They Did
WSJ filmmakers—technologist Joanna Stern and audio producer Jarrard Cole—built a ~3‑minute “robot movie” for around $1,000, using roughly 1,000 AI-generated clips
businessinsider.com
+15
bgr.com
+15
dicloak.com
+15
.

Their creative stack included:

Midjourney for character visuals (the mad‑scientist Optimax 5000).

Runway tools (Act‑One, References, ReStyle) for gestures, consistency, and motion
bgr.com
+3
giacomoiotti.me
+3
dicloak.com
+3
.

Google Veo 2 initially, then Veo 3 for animating visuals with synced environmental sounds and some dialogue
businessinsider.com
+8
giacomoiotti.me
+8
theverge.com
+8
.

Suno for background music.

ElevenLabs for character voices; Joanna recorded her own lines since the AI didn’t capture her inflection perfectly
blog.google
+13
giacomoiotti.me
+13
buchman.co.il
+13
.

Adobe Premiere Pro to stitch everything together
arstechnica.com
+2
giacomoiotti.me
+2
bgr.com
+2
.

🤖 Key Takeaways
AI accelerates production and slashes costs, but it doesn’t replace creators. All the scriptwriting, key scenes, and character direction came from humans, not AI .

Scene continuity remains a headache. AI models like Veo and Runway still struggle maintaining consistent characters across clips, requiring prompt engineering and manual editing
businessinsider.com
+15
bgr.com
+15
buchman.co.il
+15
.

Veo 3’s big breakthrough is native audio—it now generates not just visuals, but also synchronized sound effects, ambient audio, and voice, eliminating much of the separate sound design workflow
en.wikipedia.org
+3
medium.com
+3
linkedin.com
+3
.

🧩 The Bigger Picture
This project is a perfect example of the current AI video editing workflow: humans orchestrating a suite of AI tools, each covering one piece of the puzzle.

The result looks impressively polished—“almost indistinguishable from a mini‑movie”—yet achieving it still demanded intensive prompt iteration and editing
bgr.com
.

⚙️ Why It Matters
Tools like Veo 3, Runway Gen‑3 Alpha, and OpenAI Sora mark a turning point in video content creation, especially with audio now handled natively
axios.com
+2
axios.com
+2
en.wikipedia.org
+2
.

But underlying issues—like character consistency, voice realism, and cost of heavy prompting and editing—remain top challenges.

Overall, the WSJ experiment was wild and eye‑opening: you can now make a mini‑film for a few hundred dollars, but you still need humans in the driver’s seat for vision, cohesion, and polish.