Depth StudioHow to Make Realistic AI Characters: Put Your Character Into Real Video
Updated: October 9, 2026 · Depth Studio
AI characters now look flawless in a single frame — but once they move, most still feel "artificial". The reason is usually not the character, but that everything around it is synthetic too. This guide shows how to close that gap by combining the character with real footage.
The Depth Studio tools that do this
Kagami — one video, a new cast
Keep the motion and the scene of any clip and recast it with your character — or swap one thing in the shot for your photo. The original sound stays.
Try Kagami →Scene Compositor — everything in one frame
Bring the character, the location, the wardrobe and the props into a single shot; say what each photo is for, then set the camera and the light. The light and the perspective settle themselves.
Try Scene Compositor →Why does real footage make an AI character more realistic?
Shrinking what AI has to invent is the most reliable way to make the result look real. In a real shot, these are already right:
- The light is real: shadows, reflections and the light on the skin all come from the same source.
- The camera is real: handheld shake, focus breathing and the character of the lens are kept.
- The body language is real: walk, weight shift and hand gestures come from a real person — the area AI struggles with most.
- The location and physics are real: wind, fabric, water and crowds stay as they are.
AI does not have to fake any of these layers; it only recasts the person with your character. Kagami's tagline sums it up: Reality, recast.
Put your character into real video in 6 steps
Create the character in Character Studio
Describe the character with concrete traits, generate a multi-angle character sheet, pick a candidate and revise it. Giving Kagami several angles of the character helps the face stay the same when the head turns.
Choose or shoot the right clip
Pick a 4–30 second clip without cuts in which the person is clearly visible. A phone video is enough; it must be footage you shot or have the rights to, and the person in it must have consented.
Trim and crop the clip
Use "Trim & crop" in Kagami to choose the part and the framing to use. Anything outside the selection is not sent.
Choose the Motion transfer mode
Motion transfer keeps the motion and the scene of the video and recasts the person with your character: the same shot, someone else performing it. If there are several people in frame, write which one should change (e.g. "the woman in the red coat").
Optionally add an outfit, product or location
Each image you add gets a role: Character, Outfit, Product or Location. The Objects swap mode replaces one thing in the shot with the photo you give — for example putting your brand's jacket on the actor — and keeps everything else the same.
Generate, keep the sound, review
The result matches the length of the source clip, and the video's original sound is laid back onto it. You review the output in your project history and share it with your team or client.
Shooting tips for a more realistic result
- Pick a performer with a similar build: if height, body and hair length are close to the character, the recast is cleaner.
- Keep the face clear: use frames facing the camera, in focus and with little motion blur; trim sudden turns.
- Soft, single-direction light: hard, mixed-colour lighting makes AI's job harder too.
- Use a single take: choose one continuous shot rather than an edited video with cuts.
- Match the character image's light to the video: don't give a cool studio-lit character to a warm sunset shot.
What is it used for?
- Commercials and product videos: a real location with the brand's AI face; put the product in the character's hand or on them with Objects swap. More: How to make an AI commercial.
- AI influencer content: the same face, with real motion in every video. More: How to create an AI influencer.
- Film and series pre-production: see casting, costume and location tests before the shoot day.
- Social media series: reuse motion you shot once with different characters.
Consent, responsibility and labelling
Kagami works on real footage of real people. That is why a separate agreement is confirmed before your first Kagami clip: you must hold the rights to the video you upload; every identifiable person in it must have given explicit consent; you alone are responsible for the videos you upload, the people in them and the resulting clips; and when you publish output that could pass as authentic, you disclose that it was generated or altered with AI.
The last point comes from Article 50 of the EU AI Act, in force since 2 August 2026.
Frequently asked questions
How do I add an AI character to a real video?
Create the character in Character Studio, open Kagami and upload your video, choose the Motion transfer mode and add the character image. Kagami keeps the motion and the scene of the video and puts your character in place of the person.
Is the video's sound lost?
No. Kagami generates the clip silently and lays the source video's original sound back onto the result; dialogue and ambience are kept.
How long can the videos be?
Clips of 4 to 30 seconds. From a longer video, you pick the part you want on the Trim & crop screen.
Can I use someone else's video or a celebrity?
Not without rights and consent. You must hold the rights to the video you upload and every identifiable person in it must have given explicit consent. That responsibility is yours under the agreement we ask you to confirm before your first Kagami run.
What if the result is not realistic enough?
The fix is usually in the footage: choose a sharper single-take clip with soft light, add several angles of the character, and if there are several people in frame, write which one should change.
Sources and further reading
Create your character, put it into real video
In Depth Studio you create your character in Character Studio and place it into real footage with Kagami — motion, light and sound are kept. Early access is opening in limited groups.
Request early access →