← All articles

AI video generation in business: Sora, Veo, and what's actually useful

02 September 20252 min read

Mature AI video models in 2025. For marketing, training or product: where they pay off and where not yet.

Since late 2024 AI-generated video left the "wow demo" phase. OpenAI Sora, Google Veo, Runway: they now produce 5-30s clips usable in marketing, training, prototypes. What is serious for business?

The three players

  • Sora (OpenAI): top quality on cinematic scenes. Limited: 20s max, costly, high-tier access only.
  • Veo (Google, via Gemini): excellent quality, synced audio, Google Workspace integration for business.
  • Runway Gen-3 Alpha: most mature for pro workflows (video extension, image-to-video, motion brush). Subscription.

Use cases that work today

  1. B-roll for corporate videos: landscape, generic people, objects — all generated. Saves on stock and shooting.
  2. Explainer animations: turning a screenshot into a video with camera motion. For landing pages and demos.
  3. Ad prototypes: client pitches without real shoots. Approve the concept, then shoot.
  4. Internal training: scene mock-ups for anti-phishing, safety, soft-skills courses.

What still does NOT work

  • Videos with identifiable people: deepfake/consent issues. Serious companies avoid.
  • Videos with on-screen text generated by the model: still glitchy.
  • Fast motion with complex physics: the model invents.

Real costs

For a B2B retail client we produced 12 B-roll videos of 8-15s: ~€180 total (Runway). Traditional shoot: ~€3,500 minimum. Human time: 6 hours of prompting + curation vs 2 crew-days.

Compliance

The AI Act requires labelling generated content. Practically: add metadata and (visibly) an "AI-generated" disclaimer on the first frame. For ads or info-sharing videos, mandatory.