Multimodal should still feel personal
The same character, heard and seen in more ways.
Media matters when it deepens a moment, not when it interrupts one. Palumi treats voice, images, music and video as expressions of the character relationship rather than separate generators with unrelated personalities.
One identity profile, model-specific delivery
Text style, voice direction and visual references come from the same character profile. Each provider still has its own supported voices, codecs and file limits, so Palumi maps the canonical identity to compatible settings instead of blindly sending one parameter everywhere.
When perfect voice continuity is technically unavailable, the experience should be honest. A song can be framed as a requested performance from a friend rather than quietly giving the character a completely different voice.
Native file first, link when necessary
The website can display media directly. On messaging platforms, Palumi prefers a native image, audio or video and uses a playable link only when that destination cannot accept the original.
Creation time is part of the scene
Images and video can take time. Palumi acknowledges the request immediately, then uses character-specific progress messages rather than asking the user to stare at a spinner. The final media returns to the same conversation when ready.
Last updated