The full application

Embodiment 08 · Image Reading

Your napkin sketch, read like a draftsperson.

Upload sketches, photos, or diagrams — or let batch mode pull images straight out of disclosure documents. A vision model reads each image in the context of your invention and returns what drafting actually needs: a figure description, a title, and consistency with your reference numerals.

YOUR SKETCH10HOUSING12SENSOR HEAD14CONDUITNUMBERED · CAPTIONED · CLAIM-MAPPED
Fig. E8Image Reading

4

formats read · PNG JPEG WebP SVG

(12)

descriptions aligned to your numerals

100%

analyzed before drafting begins

The method

  1. 01

    Drop in any image

    PNG, JPEG, WebP, or SVG — uploaded in the Figure Planner or embedded in batch documents. Each image is queued for analysis the moment it lands.

  2. 02

    Preprocessed properly

    Images are validated and normalized — downscaled to sane dimensions, SVGs rasterized — before any model sees them.

  3. 03

    Read in context

    The vision model receives your invention summary, existing diagrams, and declared components — so its description aligns with the reference numerals your draft already uses.

  4. 04

    Returned as a figure

    A detailed description and a suggested title populate the figure record — feeding the Brief Description of the Drawings and the Detailed Description automatically.

  5. 05

    Ready before drafting

    In the automated pipeline, image analysis completes before section drafting begins — figure text is never an afterthought.

Wherein, notably

“The round thing on the left” comes back as “sensor head (12)” — because the model read your components first.

Wherein

Context-aware vision

Because the model knows your components and numerals, “the round thing on the left” comes back as “sensor head (12)” — consistent with the rest of the draft.

Sketches and AI figures coexist

Uploaded images and AI-generated diagrams live in one figure sequence, numbered and ordered together.

A real queue behind it

Vision jobs run with heartbeats, retries, and self-healing — a hundred embedded images in a batch won’t fall over.

Transparent provenance

Each analyzed figure records the model used, its status, and any warnings — you can always see how a description came to be.

See it on your own invention.

Upload the sketch · get the figure

Start your application

Free to start · No credit card required