GlossiDocs
Start free
Create in Studio

Ask the Assistant

Change the scene, generate AI image and video edits, and ask questions about the scene in plain language.

The Assistant changes the live 3D scene from a plain-language request, hands requests the scene cannot express to AI image or video editing, and answers questions about the scene. An AI edit is a separate result: your scene and original render stay as they are.

Open the Assistant

  • Click the sparkle icon (Agent) at the bottom of the left rail.
  • Or click Edit view with AI on the viewport, in Photo or Video mode. This opens the Assistant on the current shot or clip.

Press Escape to close it. Work you started keeps running, and the icon shows a count of new updates until you come back.

What you can ask

Change the scene. Anything the 3D scene can express: the look ("add a little film grain", "warmer, softer light"), the background, the product ("select the blue colourway", "turn the shoe to face the camera"), props ("put a few rocks around it"), saved shots and clips, and how a clip moves. The change lands in the live scene, so it applies to every shot and render.

Generate an image or video edit. Anything the scene cannot contain: a beach at sunset, a hand holding the product, lifestyle context, or "make a video of it spinning".

Ask a question. "What lights are in this scene?" The Assistant reads the scene and answers.

You do not need to say which of these you mean. When a request could be either ("make it darker" could be the background or the whole frame), the Assistant asks, with options to tap.

Point at what you mean

The composer has tools above the text field:

  • Choose scene context lists your products, materials, parts, lights, props and shots. Hovering a material or part highlights it in the viewport. Picking one scopes your next request to it, shown as a Context chip you can clear.
  • Mark a spot lets you click a point in the viewport to scope the next request to it.
  • Add reference attaches one image (PNG, JPG or WebP, under 25 MB) when a photo says it better than words. You can also drag an image onto the composer.
  • The microphone dictates into the field. Your words stay editable; press Enter to send. Dictation uses your browser's speech recognition and asks for microphone access the first time.

How it answers

Each request becomes a turn in the conversation. The reply is one of these:

  • A scene change: one sentence saying what changed. Press ⌘Z to undo a scene change.
  • A question with option chips. Tap one, or type your answer.
  • An image or video hand-off: why the scene cannot do it, the prompt the Assistant will use, and a Generate image or Generate video button. When you plainly asked for an image or video, it starts without the button.
  • Suggestions: follow-up ideas that fill the composer when tapped.

An image hand-off shows its settings before you generate: Strict keeps the product's size, angle and crop, Flexible lets the scene recompose around it, Product accuracy runs extra checks on shape, logos and key details, and a quality tier (Fast, Standard, High or Max). Higher tiers show their usage, such as "2× usage".

When the next step is something you do by hand, the reply names the place as an underlined link, such as Export or the clip's Animations tab. Click it and Studio opens that panel, switching between Photo and Video if needed.

Before the Assistant removes anything (a prop, a shot, a clip, a decal), it asks you to confirm and names what will go. A removed prop cannot be restored with undo.

Use image and video results

A generated result appears as a card in the conversation:

  • Click the image or video to open it over the viewport. You can compare it with the original or save it to the library. Go back returns to the live scene.
  • Save keeps it in your render library. Until then it stays in this session.
  • Refine makes it the target of your next request. The Context chip shows Edit and its name; describe the change ("softer shadows") and the Assistant edits that result instead of the scene. Clear the chip to go back to the scene.

Image edits use the current shot in Photo mode and the selected clip in Video mode. If you changed the shot, Glossi saves it before rendering the source. To repeat an edit on every future render of a shot, save it as instructions on that shot: see Shot settings.

Write a good request

  • Say what should change and what should stay: "Place the shoe on wet sand at golden hour. Keep its position and logo exactly as they are."
  • One change per request is easier to judge and to undo.
  • Name what you see ("the key light", "the blue variant"), or use Choose scene context to point at it.

What the Assistant does not do

  • Render or export. Use Export in the header. See Render images and video.
  • Guarantee product fidelity in AI edits. An edit can drift from the product. Compare before you use it; when exactness matters, change the scene and render again.

On this page