← Blog

Best AI to Replace Character in Video: 4 Kinds, 1 Real Test

The best AI to replace a character in a video depends on what you want to keep. A face swap keeps your footage and changes only the face. Motion transfer keeps a movement and gives it to a character image. A whole-video re-render keeps the shots, cuts and timing and changes who is in them, and a node workflow keeps whatever you build by hand. A disclosure first: we make Riffkit, which does the third kind.

The car scene side by side: the original woman on the left, our saved character in her place on the right, in the same gown and pose

Click for the silent side-by-side: our own 12-second clip on the left, the same clip with a saved character in her place on the right. Seedance 2.5 at 720p, about 5 minutes to render.

The kinds of AI that replace a character in a video

Sort the tools by what you give them and what they leave alone. The kind of tool matters more than the brand.

Kind of tool What you give it What stays from your video What it is good for
Face swap Your video and a photo of a face The footage itself: body, clothes, scene and movement Changing a face and nothing else
Motion transfer (also called motion control or animate) One image of a character and a video to drive it The movement and the expressions Making a still character perform a move
Whole-video re-render with a character reference The whole video and a saved character The shots, cuts, camera moves and timing. Every frame is drawn again A finished video with several shots that you want back with your character in it
Manual compositing or a node workflow The video, plus your own masks, models and settings Whatever you choose to keep Full control over every step, when you want to build the pipeline yourself

Examples. Each named tool is described from its own page as of October 10, 2026:

  • Face swap. A video face swap takes your clip and one photo of a face. It puts that face on the character in the clip and leaves the body, the clothes, the scene and the movement as they were.
  • Motion transfer. Kling Motion Control, per its guide, assigns motion from an uploaded video to one character in an image. The action video runs 3 to 30 seconds and "must be a single continuous shot, with the character consistently visible in the frame." Wan-Animate, from Alibaba's Tongyi Lab, animates a character image from a performer's video. Its page describes a second mode that puts the animated character into the reference video "to replace the original character", matching the scene's lighting and color tone.
  • Whole-video re-render. This is what Riffkit's Swap mode does, for a clip of up to 45 seconds. Higgsfield's Genjutsu page says its Motion Transfer mode "keeps the motion, camera and timing while rebuilding the scene from your references", for reference videos of 4 to 30 seconds.
  • Node workflow. ComfyUI's docs call it "a node-based interface and inference engine for generative AI" that is open source and can run on your local device. You connect the models and the steps yourself.

We set Kling Motion Control next to Swap, row by row, in our Kling Motion Control comparison.

How to choose

  • You want the same shots and cuts with a different character: re-render the whole video. The source's edit is the brief. A video with several shots comes back as one video, cut for cut.
  • You have only a picture and want it to move: use motion transfer. You bring the character image and a clip of the move.
  • You want the footage left as it is: use a face swap or a replace mode. A face swap changes the face only. A replace mode, like the one on Wan-Animate's page, puts a new character into the scene you shot.
  • You want the same character across many videos: save the character once. The tool then works from the same photo every time. Our guide to keeping one AI character across every video explains why that holds a face steady.

Whichever you pick, use your own face, a character you created, or one you have the rights to. The same goes for the video you start from.

How to put yourself in a video with AI

To put yourself in a video, save your own photo as a character once, then pick the route by what you start from:

  • A video you already have, or a template: Swap. It keeps the shots, cuts, camera moves and timing and puts you in the lead. Your face, hair and build come from your photo, so more than the face changes. To keep the person and change only the hair, see how to change hair color or hairstyle in a video. The clothes stay the original's unless you name new ones, though a scene can pick up the outfit from your photo, as one did in our test below.
  • A video that's already winning, made your own: Adapt. It keeps the winner's hook, rhythm and beats and makes new footage with you in it, in a new story, in any of 10 languages in the web app. This is the route for another creator's video: the formula carries over, their footage doesn't.
  • Only an idea: Creation. Write what should happen, pick your character, and Riffkit writes and renders an original video of up to 45 seconds with you in it.

The photo. One photo of you alone, facing the camera, your face large in the frame, on a plain background. The layout we recommend is one image with a chest-up close-up on the left and you head to toe on the right, in the same outfit, so the model sees your build as well as your face. The photo passes the video platform's review before its first use, usually in under a minute.

Your voice. Add a 4 to 15 second clip of you speaking to the character, as an mp3 or wav with no music behind it (5 to 10 seconds works best). Adapt and Creation then use it as your voice on spoken lines. In Swap, the voice follows the engine, as the steps further down explain.

Consent. If it's your own face, using the feature counts as your consent. Anyone else who appears in your video, including a person a Swap keeps from the original, must have given you their written consent first, and no one under 18 can be shown.

A free account can try all of this with one character and one photo, at 480p; a video with more than one section gets its first section free, and a plan finishes the rest. Our test below used a saved character; your own photo goes through the same steps.

A real test: one saved character, one 12-second clip

We ran the third kind on a video of our own. Here is what came back.

The source is our own 12-second clip, with scenes on a fire escape, in a car and on a red carpet, and a caption on screen. In Swap we picked a saved character and wrote one line that only named her. The engine was Seedance 2.5 at 720p, and the render took about 5 minutes.

What stayed. Her face is in every shot. The shots, the cuts and the caption stayed where the original has them. In the car and on the red carpet she wears the original's gown and gloves.

What drifted. One scene. On the fire escape she wears the jacket from her character image, and she holds the phone in her hand where the original wedges it between shoulder and ear.

The rerun. We ran it again at 480p with the "What to change" box empty. Her face was in every shot that time too.

So her face was in every shot, and one scene picked up the jacket from the character image. The whole video is rendered again, so small things can move, and it pays to watch every scene before you post. Our Seedance character swap guide has the photo layout we recommend and the rate for each engine.

How to replace a character with Riffkit Swap

  1. Create the character once. In the web app, open Characters and tap + New character. Add a name, a gender, an age range and one photo: a single face, front-facing and large in the frame, on a plain background. The layout we recommend is one image with two views: a close-up of the face on the left and the full body on the right. The photo passes the video platform's review before the character's first use, usually in under a minute, and the status turns Ready.
  2. Open Swap.
  3. Add your video or pick a template. A video can be an upload of up to 100 MB or a TikTok link of your own post. Swap keeps the full length, up to 45 seconds, so trim a longer clip first.
  4. Pick your saved character. Each character makes its own video.
  5. Optionally, write under "What to change". Name a prop, an outfit or the setting. Leave it empty and only the character changes. To change the clothes as well, our outfit swap test has the exact brief we used, and the garment-photo test does it from a picture of the clothes; to move the scene too, how to change the background of a video.
  6. See the price. The page shows the rate per second, then the billed seconds and the credit total, before you submit. When a free account starts with the first section, it shows that section's credits and your balance. For a new upload or link it is an estimate until Riffkit has the file.
  7. Tap Riff it →. On the Seedance engines, each part of a new clip first passes the video platform's content review, which can take several minutes. A refused clip costs nothing.

The voice depends on the engine. On Seedance 2.0 and 2.0 Fast, the original voice stays. On Seedance 2.5 and MiniMax H3, the new character says their on-camera lines in a new AI voice, and narration keeps the original voice unless you ask for a change in "What to change". Writing "keep the original audio" keeps every voice. Otherwise, where your character speaks on camera, or your text changes a voice, that part's sound is newly rendered on every engine: the voice over the source's music track when Riffkit has one, and the source's effects and other voices drop out there. Everywhere else the sound is the source's own.

Working in a chat instead? The same swap runs from Claude, or from ChatGPT, where Riffkit is in the plugin directory, and replace the person in a video has the sentence to paste. How to make videos in ChatGPT and Claude has the setup.

What you can do for free

You can swap a character in a video on a free account. Signing up takes no card.

  • Credits. New accounts start with signup credits, and the page shows what a swap costs before you spend them.
  • Engines. Free accounts render at 480p, on Seedance 2.0 Fast (the default) or Seedance 2.0.
  • Length. A short clip comes back whole. A clip with more than one section gets its first section on the free account, and a plan finishes the rest.
  • Watermark. Free videos carry a small Riffkit watermark. Any plan removes it from every video, earlier ones included.
  • Characters. A free account keeps one saved character.

Seedance 2.5, MiniMax H3, 720p and 1080p come with a plan: Lite is $15 a month for 3,000 credits, Creator is $39 for 8,000 and Studio is $99 for 22,000. Swap bills per second of finished video, from 50 credits a second on Seedance 2.0 Fast at 480p, and failed renders aren't charged.

Limits to know before you start

  • Small details can drift between shots. A jacket or the way a hand holds a phone can change, as both did in our fire-escape scene. Watch the whole result before you post.
  • A character needs a photo behind it. A character described only in words has no reference image, so the face can differ between shots. A saved character gives the model a photo to work from in every shot. In our words-only test, a new woman described in words got the new haircut but kept the original face.
  • The source needs someone to replace. A character counts as a change only when the video shows someone on screen. A swap with nothing to change stops before anything is billed.
  • Use it on what is yours. Swap is for your own videos, templates, and characters you have the rights to. It keeps the source's sound wherever no voice changes, and someone else's video still belongs to them; a template's music, voices and people aren't rights-cleared either. For another creator's winner, Adapt, Riffkit's other mode, keeps the formula and makes new footage. When Riffkit has the source's music as its own track, Adapt keeps that BGM by default: the music is part of why the video won, and the new video follows its beats. The track's rights stay with its owner, so check you're cleared to use it where you post (where the line sits).

Try it on your own clip

Set up one character, then open Swap with a clip of up to 45 seconds. The price shows before anything renders.

FAQ

What is the best AI to replace a character in a video?

It depends on what you want to keep. A face swap keeps the footage and changes only the face. Motion transfer tools, such as Kling Motion Control, make a character image perform the movement from a reference video. A whole-video re-render keeps the shots, cuts, camera moves and timing and changes who is in them. Riffkit's Swap mode is a whole-video re-render for clips of up to 45 seconds: you pick a saved character, and the video is rendered again with that character in the lead. In Riffkit's own 12-second test on Seedance 2.5 at 720p, the saved character's face was in every shot, the shots, cuts and caption stayed, and one scene picked up the jacket from the character image.

Can I replace a person in a video with a photo?

Yes. In Riffkit, save the photo as a character first: open Characters, tap + New character, add a name, a gender, an age range and one photo of a single person, front-facing with the face large, on a plain background. The layout Riffkit recommends is one image with a close-up of the face on the left and the full body on the right, in the same outfit. The photo goes through the video platform's review, usually in under a minute; if it reads Review failed, upload a different photo. When the character reads Ready, open Swap, add your video and pick it: the shots, cuts and timing stay, and the person changes. Use your own photo or one you have permission to use.

How do I put myself in an AI video?

Save your own photo as a character first: one photo of you alone, facing the camera with your face large, on a plain background, or one image with a chest-up close-up on the left and you head to toe on the right. Then pick the route by what you start from. In Riffkit, Swap keeps the shots, cuts and timing of a video you have and puts you in the lead, taking your face, hair and build from the photo. Adapt keeps a winning video's hook, rhythm and beats and makes new footage with you in it. Creation makes an original video of up to 45 seconds from an idea you write. A free account can try it with one character at 480p, and a video with more than one section gets its first section free.

Can I put myself into a trending video with AI?

Yes. For your own video or a Riffkit template, Swap keeps every shot and puts you in the lead. For another creator's trending video, Adapt keeps its hook, rhythm and beats and makes new footage with you in it, so your video works like the winner without reusing their footage. Both start from one saved photo of you. If it's your own face, using the feature counts as your consent; anyone else who appears needs to have given written consent.

Can I replace a character in a video with AI for free?

You can try it free in Riffkit's Swap mode, with no card needed. New accounts start with signup credits and can swap at 480p on Seedance 2.0 Fast or Seedance 2.0, with one saved character. A short clip comes back whole. A clip with more than one section gets its first section on the free account, and a plan finishes the rest. Free videos carry a small Riffkit watermark, and any plan removes it from every video. Seedance 2.5 and 720p come with a plan, from $15 a month for 3,000 credits.

Does the replaced character keep the original voice?

It depends on the tool and the engine. In Riffkit's Swap mode, Seedance 2.0 and 2.0 Fast keep the original voice. On Seedance 2.5 and MiniMax H3, the new character says their on-camera lines in a new AI voice, and narration keeps the original voice unless you ask for a change in "What to change". Writing "keep the original audio" keeps every voice. Otherwise, where your character speaks on camera, or your text changes a voice, that part's sound is newly rendered on every engine: the voice over the source's music track when Riffkit has one, and the source's effects and other voices drop out there. Everywhere else the sound is the source's own. To use your own voice, add a 4 to 15 second voice sample to your character: Adapt and Creation use it on spoken lines; in Swap, the voice follows the engine.

How long can a video be when you replace a character with AI?

It depends on the tool. Riffkit's Swap mode takes a clip of up to 45 seconds and keeps its full length, so trim a longer clip to 45 seconds first. As of October 10, 2026, Kling's Motion Control guide lists action videos of 3 to 30 seconds in a single continuous shot, and Higgsfield's Genjutsu page lists reference videos of 4 to 30 seconds.

Keep reading

Add an Object to a Video with AI (Real Test)

Add an object to a video with AI and keep the person, moves and sound. Our test put sunglasses and a plant into an 11-second clip, checked every half second.

Make a Christmas AI Video from a Clip You Already Have

Make a Christmas AI video from a clip you already have: four Seedance 2.0 tests with lights, a tree, pajamas and snow, the exact prompts, and what slipped.

Change the Person in a Video with AI (Real Test)

Change the person in a video with AI: eight tools compared, and our 11-second test where words changed her hair but not her face. Why a photo helps.