Using elements (and your model) in a prompt

In tools that take a written prompt, you can drop in reusable elements — saved characters, products, styles, backgrounds, and more — as well as your portrait/voice model. Each one becomes a colored chip inside your prompt so the AI knows exactly which saved thing you mean.

An element is a saved, reusable building block (a specific character, product, outfit, style, background, and so on). When you reference one in a prompt, it shows up as a small colored pill with the element's name and thumbnail, sitting inline with your text. That tells the model "use this exact element here" instead of guessing from words alone. For everything about creating and managing elements, see Elements.

Open the element picker from the prompt — labeled Select influencer / element — and tap the one you want. It's inserted as a chip right where you're writing. You can add as many as the tool allows, mixing them with normal text to describe your scene.

The picker groups choices into two sections:

If a tool has no model or voice to offer, the Influencer section is hidden and you'll just see your elements.

Delete its chip from the prompt the same way you'd delete a word — the reference goes away and the model no longer uses that element. Adding the same element again re-inserts the chip.

Each chip is a direct pointer to a saved image Element, so the model pulls that Element's look from its reference images into the result wherever you placed the reference. Putting a character chip and a product chip in the same prompt, for instance, tells the model to feature both in the scene you described.

Yes — it depends on the tool and the quality tier you've picked. Where a limit applies you'll see "You can add up to N elements," and trying to add more than that won't work. Many flows tie element availability to quality: you may see "Increase quality to unlock more elements" or a note that an element is only available at certain quality tiers, so raising the quality can unlock more slots. The exact count is shown in the in-app message for your current mode and tier.

A few cases are blocked with a message:

Yes. Each model, voice, and element chip gets its own consistent color based on its name, so the same element always looks the same across your prompts and pickers.

The same two input families used on desktop also apply on mobile. Generate Image and Text to Video are prompt-integrated: their element pills stay inside the mobile prompt editor. Image to Video, Chat to edit, Edit everything, and the categorized Create storyboard flow use standalone element panels.

Chat to edit and Edit everything use the non-categorized expandable mobile element field instead of the desktop-style pink header. Chat to edit's grey close bar is labelled Element and additional image, while Edit everything's is labelled Elements. While their pools are empty, the closed + Element shortcuts open the picker directly and expand the shared panel only after an Element is saved; cancelling leaves it closed. Edit image's + Additional image likewise opens the native file chooser directly and expands the panel after files are selected. Once either panel has any reference content, all of its closed-state shortcuts only reopen it. Image to Video uses the same grey-bar interaction inside a categorized Element / Voice mode / Camera field embedded in each Description card: an empty pool lets each footer shortcut open its picker, while any existing reference makes every shortcut reopen the field. These external closed-state shortcuts are consistently 24px high. Voice mode shares one selected model and one selection drawer with the footer shortcut; its expanded card fills the row. Compact chips inside visible panels remain 22px high, and selected Element and Voice model chips use a larger 24px thumbnail without changing the full-width cards elsewhere. Empty Element chips use +; selected Element chips use @ before the thumbnail in every mobile flow, including the prompt-integrated Text to Video and Generate Image inputs. The compact Element category shows the available slots and filled elements. Across these flows and Create storyboard, the handle matches Text to Video's 44 × 20 compact chevron and opens the full list as a height-capped, internally scrolling overlay floating over the lower portion of its own editor card rather than being clipped inside it or appearing above the handle. In quality-capped flows, it shows the current quality's full allowance plus one locked upgrade slot when a higher quality allows more. Add element opens the standard create/pick flow, tapping a filled element inserts its chip at the current cursor, the info control opens its details, and the cross removes it after confirmation. Create storyboard keeps its own category-specific slots while reusing the same expandable-panel interaction.

Generate Image and Text to Video both show the current quality's full allowance plus one locked inline pill whenever a higher quality unlocks more. On mobile, tapping that locked pill shows its unlocking-quality hint as a toast, switches automatically to the lowest quality that unlocks it, and reveals the new allowance plus the next locked pill; their shared card popup scrolls vertically when needed. In multi-shot Text to Video, each shot has its own pill row. In multi-shot Image to Video, opening the categorized Element / Voice mode / Camera field from any shot shows it in every shot. Saved element slots and the selected voice model belong to the whole generation, while Camera selection and Element or Voice mode insertion still target only the shot where you use the control.

The picker opens as a full-screen drawer titled Select influencer / element, sliding in from the right. It lists the Influencer rows first (your Portrait Model and Voice model, when present), then your Element rows underneath, each with a round thumbnail, name, category tag, and description. Tap a row to insert it; tap the close icon to back out without adding anything.

Related: Elements · Choosing the model · Picking a source image or video · Voice models