Analysis of Award-Winning Works from Runway AI Video Contest

Last month, Runway held a 48-hour AI video contest, requiring that 75% of the content be produced using Runway, with scene and character options provided by the official guidelines. Recently, eight award-winning works were selected from tens of thousands of entries. Today, I will analyze these award-winning works one by one.

All video subtitles were translated using Tencent’s AI video subtitle translation tool, so there may be some inaccuracies to bear with.

A few days ago, Runway released the upgraded version of gen2, which has significantly improved video clarity and stability. Prior to this, Pika also previewed their version 2.0, which similarly enhances clarity and stability, but it has not yet been released.

Compared to AI painting, AI video is a whole dimension more complex. It is not just about making images move; a good film requires an engaging story, storyboarding, music, and voiceovers. The first-place award-winning work in this video contest cleverly avoided the shortcomings of AI-generated videos, using minimal generative technology in its production process. The smoothness of the characters’ movements was achieved through GEN-1 style transfer, and the voiceover was mostly done by real voice actors, resulting in a relatively low level of AI involvement.

In the past year, most AI videos we’ve seen are those that make it obvious they are AI-generated. These works are often marketing efforts by brands trying to ride the wave of AIGC hype. The moment that can truly change the industry has not yet arrived.

Whether it’s AI video or AI painting, we can only call it ‘replacement’ when the technology can achieve the same level as previous workflows and manual work; otherwise, it is just a fleeting trend.

This AI video contest has already begun to showcase a level comparable to conventional films, although there is still a lot of room for improvement.When one day you see a video or movie and can no longer distinguish whether it is AI-generated or manually produced, that will be the moment AI video is truly revolutionized.

Most of the works this time were produced using the pre-upgrade GEN2 or GEN1. If the upgraded quality had been used, the results would likely have been even better. Moreover, the tools they used were mostly limited to Runway; if Runway could be more open, there might be even more outstanding works.

The best works were mostly done with real voiceovers; the AI-generated narration indeed has a significant gap compared to real voices. However, OpenAI’s multimodal voice might have found a solution. Currently, generating emotional voices is still a blank space; if solved, the voiceover industry in film and television could undergo major changes, as many star performances rely heavily on voice actors.

Grand Prize: Dan Hamill and Jeff Wood’s “2026”

AI Video: GEN-1 Real Video Style Transfer
Voiceover: Real Voiceover

The first prize work cleverly used helmets to avoid some issues with facial lip-syncing. It is said to have been generated using GEN1. This likely involved using GEN1 video-to-video, taking real footage of actors and using AI to generate a new visual style. If it were purely based on GEN2’s image generation, it would be nearly impossible to achieve such smooth movements.

Silver Prize: Hilario Abad, Javi Leria, and Esteban Diba’s “Myth”

AI Video: GEN-2
Voiceover: Real Voiceover

Actually, I prefer this second-place work, which presents the human experience from an alien’s perspective. The control over visuals and pacing is very mature. This short film does not feature overly complex or flashy visuals, but the immersion is very strong. Thus, while AI can solve visual production issues, the ability to tell a story remains crucial.

Bronze Prize: Shang Zhu’s “Nine Dragons Night Fog” Trailer

AI Video: GEN-2
Voiceover: Suspected AI Voiceover

This was definitely created by a Chinese person, as Andy Lau appears in it, blending science fiction, gunfights, and family elements in a commercial film style. The storyboarding is excellent, and with higher clarity, it could rival Hollywood trailers. Large scenes were used to avoid matching character movements and lip-syncing.

Excellence Award: Emad Khalid Mughal’s “Hide and Seek”

AI Video: GEN-2
Voiceover: Real Voiceover
I didn’t quite understand the story, but it seems to be about future humans choosing whether to die and facing their inner fears. πŸ˜‚ The narration is very emotional and engaging. The visuals employ a lot of zooming in and out, creating a strong contrast between large scenes and extreme close-ups.
Excellence Award: Andrea Aureli and Gemma Costa’s “Chasing the Sun”
AI Video: GEN-2
Voiceover: Real Voiceover
This is an anti-war film about a girl who has experienced war chasing the sun in her car. It’s quite simple, but feels very much like a car commercial. πŸ˜‚

Best Character: Gab Michael’s “Let’s Explore”

AI Video: GEN-2
Voiceover: Real Voiceover
Indeed, cuteness is justice, even if voiced by an old man with a croaky voice. This piece is super short, feeling like it ended without telling much.

Best Production: Seif Abdalla’s “At the Gates of the City”

AI Video: GEN-2
Voiceover: Only Music
Also an anti-war film, it consists entirely of empty shots. The visuals are beautiful, but it feels lacking in focus; three minutes seem particularly long. I wonder if it’s describing the Israel-Palestine conflict; capturing the essence of empty shots is indeed challenging without dialogue or sound effects.

Best Narrative: Daniel Scott Cribbs’ “Balloon”

AI Video: GEN-2
Voiceover: Only Music
The story is about an elderly person in a nursing home reflecting on past moments with a deceased loved one, which is heartwarming. However, both the storyboarding and music are quite average. Still, it can be considered a complete story.
This AI video contest has proven one thing: only stories can touch people’s hearts. Visuals, animation, voiceovers, and music all serve the story. Simply having beautiful visuals cannot change anything.The past, present, and future of video only belong to those who can tell stories. AI merely helps us filter out these storytellers.

Leave a Comment