How to create images and videos with multiple characters with AI
Create a three-character café scene in Affogato. Follow the Studio layout, character selection, exact prompts and settings used for our image and video examples.
Three friends meet at a café in Paris. Ingrid tells a story, Marisa laughs, and Marco sets down his coffee as the camera moves closer. Here’s the ten-second scene we generated in Affogato, followed by the image version of the same idea.
Watch the full video with audio. Made in Studio with Seedance 2.5, three selected characters, and the prompt below.
What makes a multi-character scene work?
A portrait gives you one person to direct. Add two more, and you need to decide who is speaking, who is listening, and how they relate to each other. That’s what makes a group scene feel like a moment rather than a lineup.
In Affogato, you select the characters before generating, then use their tags in your prompt. We’ll recreate this café example with Marisa, Ingrid and Marco, using the same cast for an image and a video.
1. Find your way around Studio
Open Image Studio. The left side holds your generation controls; the canvas on the right holds your results.
Model and settings: choose your image model, aspect ratio and output count. Our image uses Affogato Realism, 16:9 and one output.
Characters: add the people who belong in this scene. Their names and prompt tags appear together.
Prompt: describe what each person does, followed by the setting, framing and light.
Canvas: open your results, compare attempts and download the ones you want to keep.
For video, switch to Video at the top. The available controls change with the model and mode, so check them again before generating.
2. Select your characters
Click Add another in Characters, then choose each person from the picker. For this example, Marisa is @Character1, Ingrid is @Character2, and Marco is @Character3. Follow the tags shown in your own selection; they identify who your instructions apply to.
The picker supports name search. Here it’s filtered to “Mar”; the counter includes Ingrid, selected earlier.
Choose clear character references with visible faces and distinguishing features. If you need to create someone first, start in Characters, save them, and return to Studio.
3. Write a scene you can picture
“Three friends having coffee” leaves most of the decisions to the model. Give each person a position, an outfit and a small action. Here’s the exact prompt used for our image:
A candid editorial photograph of exactly three friends at a café in Paris. @Character1 sits on the left in a cream linen shirt, laughing as @Character2, at the center in a blue cardigan, tells a story with one small hand gesture. @Character3 sits on the right in an olive jacket, smiling as he sets down a white coffee cup. All three faces are clearly visible. Warm morning light, an eye-level medium-wide composition, natural expressions and realistic skin. Keep their facial identities distinct. No other people in the foreground, no text or logos.
The positions keep the cast easy to follow. The hand gesture, laughter and coffee cup give everyone something to do. “Eye-level medium-wide” asks for enough space to see all three faces. The character references supply their identities; the prompt directs this particular scene.
4. Turn the idea into a video
In Video Studio, we selected Seedance 2.5 in Text to Video, kept the same three characters, and chose 10 seconds, 16:9, 720p, with audio on. This was a separate generation using the character references, rather than an animation of the image above.
For motion, describe a short sequence and one camera movement:
A natural ten-second video of exactly three friends at a café in Paris. @Character1 sits on the left in a cream linen shirt. @Character2 sits in the center in a blue cardigan. @Character3 sits on the right in an olive jacket. Begin with an eye-level medium-wide shot showing all three faces. @Character2 tells a story with a small hand gesture; @Character1 raises an eyebrow, then laughs; @Character3 smiles and sets down a white coffee cup. Slowly move the camera closer as they react. Warm morning light and quiet café ambience with light laughter, no spoken dialogue. Keep their faces and clothing consistent. No additional foreground people, text or logos.
Ten seconds gives this conversation room to breathe. Asking for several locations, outfit changes and camera cuts in the same clip would make this simple moment harder to follow.
5. Review the result before adding more
Check the faces first: did each person stay distinct, and did anyone change appearance as they moved? Then watch the interactions. Look at hands around cups, eye lines and whether the camera keeps the important reactions visible.
Our image includes café signage despite the request for no text. That’s a useful reminder to inspect the background, too. A prompt guides generation; it doesn’t guarantee every detail. Change one specific instruction when retrying so you can judge what helped.
A few dos and don’ts
Do assign actions by tag. Say who laughs, who looks away and who holds the cup.
Do describe clothing when it matters. Use colors and garments to make the scene readable.
Do start with a manageable moment. A conversation or shared reaction is easier to assess than a crowded action sequence.
Don’t rely on names alone. Use the character tags displayed in Studio.
Don’t give conflicting directions. A close-up of one face can’t also show everyone’s full body.
Don’t judge identity from one frame. Watch the full clip, especially turns and close-ups.
For more help, read our guides to keeping a character’s face consistent and changing outfits while keeping the same person.
Can I reuse these characters in another scene?
Yes. Select the same saved characters, check their tags, then describe the next setting and interaction. Try moving this café conversation to a podcast table or a beach picnic. Keep the cast familiar while giving them a new reason to be together.
