top of page
The textBOX Describe logo.

beautiful image descriptions,
for every reader

Describe is textBOX's new AI-assisted image description service: alt text and long descriptions, matched to what each image needs, for books, journals, and websites. At the heart of Describe lies the Inversions Library: over 200 expert methods, one for each kind of image, distilled from years of describing content for our clients. It's unique to textBOX. Every description is grounded in the text around it and signed off by a specialist, so the quality holds from the first figure to the last.

An 1857 woodblock print by Utagawa Hiroshige, Sudden Shower over Shin-Ōhashi Bridge and Atake, showing figures crossing a bridge in driving rain.

ALT TEXT (DETAILED)

​An 1857 colour woodblock print by Utagawa Hiroshige, Sudden Shower over Shin-Ōhashi Bridge and Atake, from One Hundred Famous Views of Edo. A steep wooden bridge crosses the Sumida River at a diagonal, its pale timber deck and trestle piers filling the lower half. Seven travellers hurry across in a downpour: two shelter under dark umbrellas and one under a conical straw hat, three huddle beneath a single large umbrella, and one holds a straw mat over his head. Driving rain is rendered as fine grey and black slanting lines across the whole sheet. On the green-grey water below, a lone boatman poles a timber raft. The far bank dissolves into a dark band of trees and houses beneath a black storm sky deepened by bokashi gradation. Red title and signature cartouches appear at upper right and lower left.

How it works

The Describe service is a simple process:

 

STEP 1: Send us your files: a PDF or EPUB of your book or journal, or the address of your website.

​

STEP 2: We describe and review every image, using our Describe platform, with specialist sign-off.

​

​STEP 3: You receive finished descriptions, in the format your production uses.

The Inversions Library

A chart and a painting present different descriptive challenges. A chart needs its axes, its trend, and its values. A painting needs its composition, its subject, and its mood. Treat them alike, and you get the flat, structureless description that gives automated alt text its reputation.

 

The Inversions Library, handcrafted by textBOX and based on our focus|LOCUS method, gives every kind of image its own distinct method: over 200 of them, covering charts, maps, diagrams, artworks, photographs, scientific figures, and more. Each one sets out what a reader needs to know, in what order, and at what depth. And the Library keeps evolving: what we learn from reviewing descriptions shapes how we refine it, project by project.

Written for the reader

Every description is written for the person who will hear it. It begins with what the image is and why it's there, explains how it's laid out, and then walks through what matters, in an order that makes sense by ear. Readers choose how deep to go, data comes as tables they can navigate, and every description follows the same patterns from the first figure to the last, creating a consistent user experience.

Compliance is not access

Accessibility law now expects every meaningful image to have a text alternative. Automated tools can produce one in seconds, and much of it is accurate. But accurate is not the same as useful. A description written from a generic prompt rarely tells a reader how a chart is organised, what the trend is, or why the image is there at all.

 

A good description does all three, every time. Describe, powered by the Inversions Library, is designed to create consistent, reader-focused descriptions which enable visualisation.

What Describe is

READ IN CONTEXT

Describe reads what surrounds each image. For a book, a brief of the whole book and the passage that discusses each figure. For a journal article, a brief of the paper and the section each figure supports. For a website, a summary of each page and the relevant content around each image.

THE RIGHT DESCRIPTION FOR EACH IMAGE

Every image is assessed with ECHO, our technique for judging what it needs: nothing if it's decorative, a short functional label, alt text alone, or a full long description. It is then routed to the right method and calibrated to your readers, from trade and educational titles to academic research. Calibration adjusts language, register, and tone, never the structure.

A DEFINED STRUCTURE

Long descriptions are designed in four parts: context, structure, a walkthrough or overview, and data. Values go into accessible tables, not buried in prose, so they can be explored independently by the reader.

HONEST ABOUT UNCERTAINTY

Describe is built to say what it can’t see rather than guess. The method gives it permission not to know, so there’s no pressure to invent. Anything uncertain is recorded in a confidence report, known as a Weather Report, for the describer to assess and verify.

REVIEWED AS A WHOLE

Every description passes automated house-style and quality checks, and Describe ranks the set so review starts with the figures most likely to need attention. A specialist then reviews the whole set, checks the facts, keeps related figures consistent, and signs off every description before delivery to the client, not only the ones flagged, because no writer, human or machine, catches every error they are confident about. A client’s own house style can also be incorporated into the Describe workflow.

DELIVERED READY TO USE

Descriptions arrive in the format your production workflow uses: a Word document for review, a spreadsheet with one row per image, a publisher-specific format, or EPUB markup for your typesetter, following the DAISY best practices. Website projects show each image’s new alt text beside its current one. Every delivery runs in page order, with a list of any errors we found in your source.

Describe in practice

Two images, two very different problems: a Japanese woodblock print and a data chart. Each gets its own method, and each description is written to be heard. The examples below are presented at their most detailed level to illustrate how Describe handles complex content.

Example 1

An 1857 woodblock print by Utagawa Hiroshige, Sudden Shower over Shin-Ōhashi Bridge and Atake, showing figures crossing a bridge in driving rain.

Image: Utagawa Hiroshige, Sudden Shower over Shin-Ōhashi Bridge and Atake, 1857. The Metropolitan Museum of Art, Purchase, Joseph Pulitzer Bequest, 1918. Public domain.

Inversions method: Japanese woodblock prints (Artworks)

ALT TEXT

An 1857 woodblock print by Utagawa Hiroshige, Sudden Shower over Shin-Ōhashi Bridge and Atake, showing figures crossing a bridge in driving rain.

LONG DESCRIPTION

The description is presented in 3 main sections as follows.

​​

Section 1. Context.

This is Sudden Shower over Shin-Ōhashi Bridge and Atake (Ōhashi Atake no yūdachi), from the series One Hundred Famous Views of Edo (Meisho Edo hyakkei) by Utagawa Hiroshige (1797–1858). The print dates from 1857, during the Edo period. It is a nishiki-e (full-colour woodblock print), ink and colour on paper, in vertical ōban format, approximately 34 × 24 cm. It belongs to the meisho-e genre (pictures of famous places). The print is held by the Metropolitan Museum of Art, accession number JP643.

​

The scene shows the Shin-Ōhashi (New Great Bridge) crossing the Sumida River, with the Atake district on the far bank, caught in a sudden summer downpour (yūdachi). This is among the most celebrated of Hiroshige's prints; it was later copied in oil by Vincent van Gogh.

​

Section 2. Structure.

The composition is built on a strong diagonal: the wooden bridge rises from the lower left corner to about halfway up the right-hand edge, dividing the sheet. Beyond and below the bridge lies the broad expanse of the river, and across the top a band of dark, heavy sky.

 

The whole image is overlaid by rain, printed as fine parallel lines of grey ruled across the entire sheet in two slightly different directions, creating the sense of a dense, wind-driven shower. This ruled rain is one of the print's most admired technical features.

 

The palette moves from the pale buff of the bridge timbers, through the blue-green of the river water, to the deep near-black of the storm sky. The sky shows bokashi gradation, deepening from grey at the horizon to dense black at the top. The river likewise darkens towards the lower edge of the sheet. Outlines are crisp, with the bridge's structure drawn in precise linear detail against the softer washes of water and sky.

​

Section 3. Walkthrough.

The key elements of the print are presented as follows.

​

  • The bridge and figures. The arched wooden bridge dominates the foreground, its pale deck curving up across the sheet and its trestle supports (a lattice of grey posts and cross-braces) descending into the water below. Seven travellers hurry across in the rain. At the lower left, two figures shelter under dark umbrellas, the red of their undergarments showing at the hem, while a third bends forward under a broad conical straw hat. In the middle of the span, three figures huddle together beneath a single large umbrella. At the far right, a seventh figure holds a straw mat over his head and back. All lean into the downpour.

  • The river and boat. Below and beyond the bridge stretches the blue-green water of the Sumida. At left, a long low raft or boat is poled by a single boatman, who crouches under a straw rain-cape and hat against the shower.

  • The far bank. Across the river, the Atake shore is rendered as a dark, low silhouette of trees and rooftops. A row of pale structures along the waterline (likely roofs or lumber-yard buildings) breaks the dark mass.

  • Sky. Above, the storm cloud fills the upper quarter of the sheet in dense black, from which the rain descends.

  • Inscriptions. At the upper right, a red cartouche carries the series title Meisho Edo hyakkei, beside a smaller title cartouche with a green and yellow ground giving the individual print title. Two small round censor and date seals sit in the top right margin. At the lower left, Hiroshige's signature appears in a red cartouche, with the publisher's seal in the lower left margin. The exact wording of the seals cannot be read with confidence from this image.

 

Description ends.

Example 2

A bar chart comparing the remaining CO₂ emissions budget for a 50 per cent chance of staying below 1.5, 1.7, and 2 degrees Celsius.

Image: Our World in Data, How much more CO₂ can we emit while staying below 1.5°C and 2°C?, 2026. Licensed under CC BY.

Inversions method: Bar charts (Statistical charts)

ALT TEXT

A bar chart comparing the remaining CO₂ emissions budget for a 50 per cent chance of staying below 1.5, 1.7, and 2 degrees Celsius.

LONG DESCRIPTION

The description of the bar chart is presented in 4 main sections as follows.

 

Section 1. Context.

This infographic, titled How much more CO₂ can we emit while staying below 1.5°C and 2°C?, is from Our World in Data, published on 12 November 2026 under the heading, How much CO₂ can the world emit while limiting global temperature rise? Its subtitle reads, Estimates of the number of tonnes of carbon dioxide (CO₂) the world can still emit to have a 50% likelihood of staying below each temperature level. This increase is relative to pre-industrial temperatures. The chart addresses the carbon budgets linked to the 2015 Paris Agreement, which aimed to keep global temperature rise well below 2 degrees Celsius and to limit it to 1.5 degrees Celsius.

A carbon budget is the amount of carbon dioxide the world can still emit while keeping within a given temperature limit. The chart shows these budgets for three temperature thresholds, each based on a 50 per cent likelihood of staying below that level. To give a sense of scale, each budget is expressed as the number of years it would last at current emission rates. The author's purpose is to show how little room remains, particularly for the 1.5 degree target, which would be exhausted within a few years at present rates.

​

Section 2. Structure.

The chart uses vertical bars, with a separate horizontal reference bar at the left.

The reference bar, a short horizontal bar in dark blue, represents projected 2026 emissions of 43 billion tonnes. It is labelled, 2026 emissions (projected). An arrow runs from it to the first red bar, inviting the comparison.

​

The three main bars are dark red with horizontal dashed lines running across them. Each bar represents a carbon budget for a different temperature limit. The horizontal axis is labelled with the three limits, from left to right: Below 1.5 degrees Celsius, Below 1.7 degrees Celsius, and Below 2 degrees Celsius. There is no numbered vertical axis; instead, each bar carries its value as a white label within it, and a red label above stating the equivalent number of years of current emissions.

​

A small bracket beside the first red bar marks its height in multiples of the 2026 emissions bar, with tick marks at 1x, 2x, and 3x. The dashed lines across each bar appear to mark these yearly increments of emissions.

​

Two annotation boxes point to the bars. The first, pointing to the Below 1.5 degrees Celsius bar, reads, If emissions stay at 2026 levels, we'll exhaust the remaining budget to stay below 1.5 degrees Celsius within 4 years. The second, pointing to the Below 2 degrees Celsius bar, reads, By 2051, we'll have exhausted the budget to stay below 2 degrees Celsius.

​

Section 3. Walkthrough.

The three budgets rise sharply from left to right, reflecting how much more CO₂ can be emitted for each higher temperature limit.

​

The smallest budget is for staying below 1.5 degrees Celsius, at 150 billion tonnes, equal to 3.5 years of current emissions. The middle budget, for staying below 1.7 degrees Celsius, is more than three times larger at 510 billion tonnes, or 12 years. The largest budget, for staying below 2 degrees Celsius, is 1,035 billion tonnes, or 24 years.

​

Set against the projected 2026 emissions of 43 billion tonnes, the message is that higher temperature targets allow much larger budgets, while the strictest target leaves very little remaining room.

​

Section 4. Data.

The full dataset is presented in the following table. The values shown are printed on the chart.

​​

Temperature limit
Budget (billion tonnes of CO₂)
Years of current emissions
Below 2 degrees Celsius
1,035
24
Below 1.7 degrees Celsius
510
12
Below 1.5 degrees Celsius
150
3.5

For reference, projected 2026 emissions are 43 billion tonnes.

 

Note: Exact estimates depend on factors such as changes in emissions of non-CO₂ greenhouse gases, like methane and nitrous oxide. Data sources: IPCC AR6 WG1; Forster et al. (2026); Global Carbon Project (2026). Chart by Our World in Data, licensed under CC BY.

​

Description ends.

Tailored to your house style

Tone, register, length, spelling, and depth adapt to your list and your readers.

​

  • Audience and reading level, including language learners.

  • British or American spelling.

  • Your alt-text length limits and description lengths.

  • Alt text only, or a full long description, decided image by image.

  • Deliverables in Word or spreadsheet form, in publisher-specific formats, or as EPUB markup for your typesetter, following the DAISY best practices.

  • New methods for new kinds of content, on request.

Who is Describe for?

Describe has been designed to tell the story of every image and make your content accessible to everyone.

​

Publishers of books and journals, universities, and organisations with large collections of images that need beautiful descriptions: new titles, backlists, and everything in between. Describe works across academic, trade, and educational content, from scientific figures to artworks.

Your content and AI

  • Your content is never used to train AI models.

  • You own every description we deliver.

  • We work under your NDA and will sign a data processing agreement.

  • Every description is reviewed and signed off by a specialist before delivery.

  • None of your images or text are added to the Library.

 

Our full data handling and security statement is available on request.

How much does Describe cost?

Describe is priced per image, according to what each image needs: alt text alone costs less than a full long description. Every project is quoted individually, based on the number of images, the mix of alt text and long descriptions, how complex the figures are, the formats you need, and your timings.

​

You’ll have a clear quote before any work begins. And if you’d like to see the quality first, we’ll describe a few of your figures free of charge.

​

Because AI-assisted drafting saves specialist time, Describe costs less than our fully hand-written service, with the same specialist review and sign-off. For content where a fully human approach matters most, such as children’s books, we still describe by hand.

Who are textBOX?

noun_writer_7388503_0F4F61__1200.png

textBOX are specialists in accessible image description. Since 2018 we have described images for a wide range of publishers, universities, and organisations, trained their teams, written their alt text guidelines, and run ASPIRE, the accessibility statement accreditation for publishers and digital platforms.

​

Describe builds on that experience. It’s based on the focus|LOCUS method we have developed since 2018, and it sits alongside our fully hand-written description, training, and consultancy services.

​

Ready for your beautiful image descriptions?

We'd love to hear from you. Tell us about your project, whether it's a single title, a journal, a website, or a whole backlist, and we'll talk through how Describe can help: the right depth for your images, the formats your production needs, timings, and price.

​

Would you like to see Describe on your own content first? Ask for a free sample. We'll agree a few figures from one of your titles and the easiest way to send them, then describe them for you free of charge.

​​

The textBOX Describe logo with the tagline, Describe Everything.
End of page icon.
bottom of page