AI Video Generation Just Got Good: Sora, Veo, Runway and Midjourney

Video is the fastest-moving modality

Image generation plateaued into refinement. Video is where the jumps are happening — longer clips, steadier motion, better physics and, finally, usable audio.

The current line-up

• Sora — strong on imaginative, cinematic shots and scene coherence.
• Veo — excellent prompt adherence and increasingly reliable audio.
• Runway — the editor''s choice: control tools, in-painting and motion brushes matter more than raw wow.
• Midjourney — still the best visual style engine, and now the front end of a broader creative stack.

What creators actually use it for

1. B-roll and background plates that would cost a shoot day.
2. Concept boards and pitch animatics.
3. Short social clips where 6 seconds is the whole format.
4. Style exploration before committing to production.

Prompting video well

Describe the shot, not just the subject: lens, camera move, lighting, pace and mood. "Slow dolly-in, 35mm, golden hour, shallow depth of field" changes more than another adjective on the subject.

The honest state of play

Hands, text, continuity across cuts and long dialogue are still weak points. Treat AI video as a powerful ingredient in an edit, not as a finished film.