A character your brand can reuse

Before generating anything, write down who the character is. This text gets pasted into every prompt, so it is worth getting right once:

• Silhouette and proportions, the shape people recognise from across a room.
• The face: eye shape, expression, anything distinctive.
• A palette taken from your brand colours.
• One signature item, a hat, a scarf, a badge, that appears every time.
• Personality in three words, and one thing the character never does.

Generate a batch of takes from the sheet and compare them side by side. Refine the winner with follow up prompts until it is right, then save it as an avatar for 5 credits. From then on every new image starts from that saved source, at 2 credits each: new poses, expressions and settings, the same character.

Each image and each clip is generated independently, so a character drifts unless you carry it. Reuse the canonical description almost word for word, start from the saved image, and never redescribe the character from memory. Remakes of short videos go further and pass up to nine reference frames to the model. The full list of drift fixes is in fix AI video glitches.

The presets include character voices as well as narrators, or you can clone a voice actor who has agreed to it. Characters are the easy case for lip sync, because nothing crosses the mouth and the head holds still, so a talking mascot usually works on the first render. More in AI cartoon lip sync.

A character is not tied to any one employee, so it survives people leaving, it can appear in every language through dubbing, and it never needs a shoot. Give it colours from your brand; the AI brand kit reads them from your website. Check that the design does not copy an existing character, and note that Percify does not watermark output on any plan.