Face & identity· 1 to 4

Consistent AI characters,the same face in every shot.

Characters & products keeps a face, a product, a place or a look the same from one AI video to the next. Save it once from 1 to 4 photos, give it a handle, and type @handle in any prompt in Create. Models that read reference pictures get them directly; for the rest, AITube composes a matching first frame.

Open Characters & products All apps

Photos: 1 to 4
Kinds: Character, product, place, style
Free plan: 3 saved

How it works

Three steps, price first.

An element is up to four photos and a name. When a prompt mentions it, AITube sends the photos to models that take reference pictures, or first draws an opening frame from them for models that don't and animates that frame. Either way the price is shown on Generate before anything runs.

Try it

Save it onceUpload 1 to 4 clear photos of the face, product, place or look, and give it a handle.
Mention it in CreateType @ and its handle in your prompt, or pick it as a chip.
Generate, keep it consistentEvery shot that mentions it starts from the same reference, so a series looks like one story.

Open Characters & products

Best results

Small things, a better clip.

Photos with a plain background and good light give the most faithful results.
For a product, include the label or logo side.
How many you can keep depends on your plan: 3 on Free, 10 on Pay As You Go, 20 on Creator and 100 on Pro.

Questions

Characters & products, in short answers.

How do I keep the same character across AI videos?

Save the character in Characters & products from 1 to 4 photos, then type @ and its handle in each prompt. Every shot then starts from the same reference pictures.

Does it cost extra?

Saving is free. When a model can't read reference pictures, AITube draws a first frame from them, which adds a few credits; Generate shows the full price before it runs.

Are my photos used to train AI?

No. AITube doesn't train AI on your uploads, prompts or voice recordings (Terms, section 5.4). What you send goes only to the models that make your result.