They say you can’t be everything at once. So naturally, I took that as a challenge.
With a blend of cutting-edge AI tools and a flair for the dramatic, I recently created a 1-minute-and-21-second short video called Be Who You Want To Be, a fast-paced, persona-hopping trip where I play multiple versions of myself, each living a very different life. Think adventure hero, corporate strategist, sci-fi explorer, style icon…all portrayed by one very overcommitted human (me).
This wasn’t just a bit of AI fun. It was a real test of something creators have been begging for in Gen AI video for the past two years: character consistency.
The Pain of AI Video… Until Recently
For those who’ve dabbled in AI video creation, you know the frustration. Until recently, if you tried to feature the same “character” across multiple AI-generated shots, things got… weird. Your subject might sprout a new face, change age, or switch outfits mid-scene. Not ideal when you’re aiming for coherent storytelling.
But now, with tools like Runway Gen-4 and Midjourney’s new Reference features, that pain is finally easing. We’re starting to get real, repeatable characters in generative video, without needing a film crew or a deepfake lab.
Behind the Scenes: My AI Video Stack
Here’s how I pulled off Be Who You Want To Be, with the help of a few powerful (and surprisingly accessible) AI tools:
- Scriptwriting: The initial story arc and dialogue were built using ChatGPT, short, punchy lines designed to showcase each persona with style and humor.
- Character Stills: Midjourney (see screenshot above) gave me beautiful, cinematic character stills using reference image inputs for consistency. Even better, Runway now has the ability to churn out great stills using References…and these stills can then be transformed into videos.
- Video Animation: Runway ML’s Gen-4 turned those stills into animated video clips. As noted above, the key here is their new character reference functionality, which lets you feed a base image so your character doesn’t mutate between scenes.
- Voiceover: I recorded the VO myself for authenticity, however ElevenLabs remains a trusty option if you want to use a library of voices, clone your own or generate some too.
- Sound FX: I layered in royalty-free sound effects, plus a few generated using ElevenLabs to bring each persona to life and punch up the transitions.
- Editing: Everything came together in Final Cut Pro, where I synced cuts to zooms, applied a light teal-orange grade for cohesion, and added bold, minimal captions with Futura (because captions matter, even when you’re mid-transformation).
Why It’s a Big Deal
This is more than a cool party trick.
Character reference control is a foundational shift for AI-generated storytelling. It means creators, brands, marketers, even hotels wanting to tell a story through a digital host or guide, can now do it without the uncanny, unintentional shapeshifting we’ve seen in past AI videos.
You can finally build a recognizable on-screen identity, and maintain it across shots.
Watch the Video: Be Who You Want to Be
Stay to the very end for the post-credits moment…let’s just say blending personas can get messy.
TL;DR: Want to Try This Yourself?
Here’s a fast-track recipe:
- Draft your script with ChatGPT or another creative tool.
- Design stills in Midjourney or Runway, using consistent prompts and reference images.
- Animate with Runway Gen-4 (I found Turbo often did a better job).
- Voice it yourself or explore ElevenLabs for stylized delivery.
- Edit in your preferred NLE – sync VO, zooms, SFX, and captions.
- Export for vertical and widescreen, then publish across TikTok, Reels, YouTube Shorts.
Final Thought
Gen AI video tools are growing up. What used to feel like novelty is now edging toward narrative-ready. And character consistency, once a blocker, is finally getting the fix it deserves.
So go ahead: reinvent yourself. Tell a story with your digital twin. Be who you want to be.
Just maybe don’t mix your fashion-forward persona with your moody philosopher all at once. Trust me on that one.