Skip to content

A.I. drawings

The spectator thinks of a word — and a moment later you have a drawing of it: a child-like sketch on your Doodle screen, a picture of their city, a calligram made of their own words. Show it on a peek device, print it, or reveal it on a display.

All A.I. drawings start from the current text of the Text Buffer (normally the last one), and they all end up in the same places:

  • on the Doodle screen (and inverted, white on black, on the BLACK screen),
  • in the Image Buffer — immediately, in original size (Image types to save → A.I. drawings, on by default),
  • on a connected Lumen Prism, if its Peek drawings option is on.

While the A.I. is drawing, a connected PeekSmith shows Drawing..., and the phone vibrates when the picture is ready.

What you need

The A.I. drawings need the A.I. Features add-on (included in every subscription) and an engine: your own OpenAI or Gemini API key — or, for PREMIUM subscribers, the free Gemini image generations. See A.I. recognition for getting a key.

The drawing actions

In the action picker, the Text buffer group has:

Action What it draws
text to OpenAI drawing the text, as a clumsy child-like marker drawing (OpenAI)
text to Gemini drawing the text, as a drawing (Gemini)
City to OpenAI / Gemini color drawing an iconic view of the city (in color)
City to OpenAI / Gemini B/W drawing the same, black and white
City+title to ... drawing the city view with the city's name written on it
hidden reveal - A.I. image a calligram: a picture drawn entirely from a repeated word (see below)
text to TrueSketch image / text to TrueWrite image a sketch / handwriting by TrueSketch / TrueWrite (see below)
text to Quickdraw a doodle from the offline Quick, Draw! collection (no A.I. key needed for the drawing itself)

Put them on any trigger — for example Action after text recognition: the spectator writes a word, and the drawing appears by itself. See Triggers and actions.

Elsewhere:

  • OpenAI drawing remake (Doodle screen group): the A.I. looks at the drawing on your Doodle screen and draws its own version of it.
  • execute A.I. image prompt (Special group): your own prompt, e.g. "a child's drawing of #TEXT on a beach" — see A.I. image prompts.

Send the drawing anywhere

Every A.I. drawing is in the Image Buffer at once, so the next step can be image buffer to printer, image buffer to Teleport, image buffer to Prism... — or an image template with #LAST1.

OpenAI settings

In Settings → A.I. FEATURES, the OpenAI part:

  • OpenAI API key
  • OpenAI drawing size: square (1024x1024) (default), portrait (1024x1536), landscape (1536x1024)
  • Use thicker lines for drawings: bold marker lines (not for the city drawings)
  • Action after OpenAI drawing + Trigger action: what happens when the drawing is ready (immediately is fine: the drawing is already in the Image Buffer)

OpenAI verification

OpenAI asks some accounts to verify their identity before they can make images. If yours does, MysterSmith shows "OpenAI identity verification required" with a Verify now button.

Gemini settings

  • Gemini API key
  • Gemini drawing model: Nano Banana 2 Lite, Nano Banana 2 (default), Nano Banana Pro
  • Gemini drawing size: square (1:1) (default), portrait (3:4), landscape (4:3), widescreen (16:9)
  • Use thicker lines for drawings
  • Action after Gemini drawing + Trigger action

Free Gemini image generations (PREMIUM subscribers): 10 a day, at most one per minute, with the Nano Banana 2 Lite model. The A.I. FEATURES section shows how many are left. If you also have your own Gemini key, your key is used first, and the free generations are the backup — unless you switch on Use FREE Gemini calls first.

Hidden reveal - A.I. image

A calligram: the picture of one thing, drawn entirely from the letters of a word, repeated again and again. From far away the spectator sees a picture; up close, their own word is hidden in every line.

Settings in the Hidden image reveal card (A.I. FEATURES):

  • How does it work — a short explanation in the app.
  • A.I. image engine: OpenAI or Gemini (default) — uses that engine's size, model and Action after ... drawing settings.
  • Image subject: what is drawn — default #LAST1 (the last text).
  • Source text: the word the picture is made of — default #LAST2 (the text before it).

Both fields accept the Text Buffer variables with !uc, !lc, !tc.

Hidden reveal is currently available for MysterSmith subscribers.

TrueWrite and TrueSketch

The handwriting and sketch services of Electricks — in Settings → INTEGRATIONS. They need a linked Electricks Cloud account (ACCOUNT SETTINGS → Show details → Electricks Cloud).

  • TrueWrite writes the last text in realistic handwriting.
  • TrueSketch draws a sketch of it — a different one every time (a word it can't draw is written in handwriting).

Settings: Image style (marker — black on white, or chalk — white on black), Image shape (square, landscape, portrait), Line thickness (thin, normal, bold), Translate (TrueSketch) (Performance language or off), Action after image generated.

On an Iarvel Peek Board, the actions TrueWrite to Peek Board, TrueSketch to Peek Board and Quickdraw to Peek Board draw live, line by line.

Quickdraw

text to Quickdraw draws the word with a real doodle from Google's Quick, Draw! collection (hundreds of everyday things: animals, objects, food...). The drawings are in the app, so the drawing itself works offline. If nothing matches the word, a short "No drawing found" message tells you.

Settings in Settings → INTEGRATIONS → QUICKDRAW:

  • Translate to English using A.I. (on by default) — the Quickdraw words are English. With this switch on, the spectator's word is first translated from your Performance language by the A.I. (with the Engine for translations set in A.I. FEATURES — your own key, or the free Gemini calls of the subscribers), and matched to the closest drawing — e.g. "kutya" becomes dog, and "puppy" becomes dog too. This works for text to Quickdraw and Quickdraw to Peek Board. Off: only English words that have a drawing, no A.I.
  • Action after drawing generated + Trigger action — what happens when the drawing is ready (it is already in the Image Buffer).

LifeLike Sketcher

Realistic drawings by the LifeLike Sketcher service (Settings → INTEGRATIONS → LIFELIKE SKETCHER, needs a LifeLike API key):

  • Drawing method: sharpie (default), whiteboard, chalk, crayons
  • Number of drawings: 1–4 (step between them with display next drawing)
  • Action after drawing generated (only with one drawing)

Actions (LifeLike Sketcher group): last text to LifeLike drawing, LifeLike drawing remake (your Doodle drawing, recognized by OpenAI, redrawn by LifeLike), display next drawing. Leave at least 10 seconds between two drawings.

Recognizing drawings

The other way round — the A.I. tells you what the spectator drew: Recognize drawing (A.I.), Analyze drawing (A.I.) and Analyze photo (A.I.) in the Text/image recognition group. See Text recognition and A.I. recognition.