What has changed

German scripts now assume 130 instead of 140 words per minute. Reading time and target time now fit a calm speaking pace on camera better.

Scripts can be ticked off as done: in the prompter after the target time and in the project overview in front of the title. Done scripts move to the bottom and are crossed out, in the overview, in the prompter’s script picker and on the phone. Untick it and the script jumps back to its place.

Recording markers such as [Bild 2.1], [Kapitel 3] or [L1] no longer appear while prompting, not on the iPad, the phone or the review page. They are meant for the recorder, not for you. They stay in the script and are visible in the editor.

When you edit on the iPad, the text is now always upright and unmirrored, even if the prompter runs upside down. Whoever writes stands in front of the iPad, not behind the glass. After saving, your usual view is back right away.

Pauses on the phone now look the same as on the iPad: a duration with a pause sign instead of an instruction in brackets.

The phone now has an “Edit” button. It opens the paragraph that is on the reading line, and in the sheet you can step to the previous or next paragraph. Pressing and holding a paragraph still works. The iPad takes over the correction when you save, without stopping.

When a script is saved on the iPad or in the editor, the phone shows the new wording by itself within a second or two. Before, it kept the old version until you reloaded the remote.

The link in the confirmation email now also works on another device, for example on your phone or in your mail app's browser. Before, it ended up on the sign in page there. The first click confirms the address and signs you in right away, and the page on your computer continues on its own.

The “Confirm your email” page shows which address the link went to and reminds you of the spam folder. If you register with an address that already has an account, you get the way to sign in and to reset your password right there.

A [Pause 3] now appears larger in the text and as a duration, with a pause sign and “3 s”, so you can see it coming. When it reaches the reading line, the line lights up and the Start button at the edge counts down the seconds. When following your voice, the pause still does not stop the text, your voice sets the pace; the display disappears as soon as you carry on speaking. Nothing covers the text.

The iPad stays awake. When the system dropped the screen wake lock, for example on entering full screen, the prompter only asked again on the next touch, and on set nobody touches the iPad. It now re-requests it on its own every ten seconds.

After a text has run all the way through, the prompter starts again. Until now “Start” did nothing after jumping back with “Top”.

With “Upside down” on, the remote no longer jumps to the end of the text. The prompter measured its position on the rotated screen and therefore reported the end of the text; it now measures within the text itself. The same affected chapter jumps, pauses and dragging on the phone.

Every page now has “Withdraw from contract here” at the bottom. There you can declare a withdrawal without signing in, in two steps: enter your details, check the summary, confirm. The acknowledgement with date and time arrives by email right away. The withdrawal policy is now labelled as such in the footer.

The legal notice, privacy policy, terms and cancellation policy are no longer drafts but the texts in force, with our address and the contact address [email protected]. The privacy policy now also explains what happens to your voice when the prompter follows along and what the devices on set report to our server. They are now available in English as well.

When you order, you now say whether you are ordering privately or as a business and accept the terms. Afterwards you receive a contract confirmation by e-mail with the terms and the information on your right of withdrawal attached as PDF files.

After the first year, consumers can cancel at any time with one month's notice. Whatever you have already paid for the time after that is refunded pro rata, automatically for card and direct debit.

Whatever comes up at a sentence while recording now sits as a note on that very sentence: it carries the mark “From the recorder”, one click takes you to the spot, and you tick it off like any other. The prompter shows none of it, there you only see your text.

Directions meant for the camera rather than for you now sit small and pale in the prompter: “[Bild 2.1]”, “[Kapitel 3]” and “[L1]” through “[L6]”. They stay readable when you look for them and no longer get in the way while you read. “[Pause 2]” stays as it was, that one is for you.

In the project list the whole row now opens the project, not just the name. “Open prompter” sits beside it as a button and stays clickable on its own.

In a project’s list of scripts, a quiet note on the right says when the wording was last worked on and by whom. Renaming and reordering do not count. If the change came in over a share link, it says so.

“Sign out this device” and “Sign out everywhere else” now really end the other device’s session, including “stay signed in”. Before, the wrong session id was recorded and the other device stayed signed in. Changing your password now also ends all other sessions, and a locked account ends at once rather than at its next sign-in.

A share link with the “remote only” permission could still read script bodies over the wire. And the prompter showed viewers the token of a share link they would never have seen in the list. Both are closed.

“Forgot password” now answers the same way whether or not the address exists. Registration and password requests are rate-limited, and the operator’s “registration open” switch now takes effect. User text in emails (comments, account names) is no longer interpreted as formatting.

A paid invoice can no longer be cancelled, only balanced with a credit note: its own line with its own number and a negative amount, as a PDF like the invoice. In your invoice list it appears as “Credit note for …”.

A tap on the phone now reaches the prompter faster, and the confirmation comes back faster: as soon as something happens, both devices briefly poll four times a second, and the prompter reports its position right after acting instead of waiting for the next tick. “Top” followed by “Start” are now a quarter of a second apart, not up to a full one. What you tap on the iPad itself (start, script change, blackout) shows on the phone right away as well.

The text on the phone trailed the iPad by about a word at all times, because it counted from when a report arrived rather than from when it was made. And during a [Pause 3] it kept moving on the phone and was then pulled back; now it stands still there too.

The second prompter following the first no longer re-measures the whole script on every frame. On an older iPad that was the single most expensive thing it did, for nothing.

Two versions now stand side by side across the full width of the screen, with the differences in colour: changed in yellow, new in green, dropped in red, word by word inside a paragraph. Which two is up to you: any version against any other, or one against the current text, and that is the one in the editor, unsaved changes included. You get there with “Compare” on any version.

A single version can be deleted. That is final, so it asks first; your current text is untouched by it, because it does not hang on any version.

The model is no longer set on the connection but chosen in the revise dialog: that is where you know what you are up to, and spelling deserves a different model than a translation. The choice is your provider’s model list. It arrives on its own as soon as it is clear whose it is: with one key when you open the dialog, with several as soon as you have picked one. Sorted into “probably a fit” and “everything else”; nothing is left out. Next to it sits a thinking level (low, medium, high), which on standard is not sent at all, because not every model knows one. Whatever you used last time is proposed the next.

There are no empty fields below your connections any more. The form for a new connection only shows up while you have none, or when you press “New connection”.

While your AI is thinking, an endless bar runs, with the words “Your AI is working” and the seconds next to it. With a big model it can take a minute until the first sentence arrives, and until then the window looked as if it were doing nothing.

The header has been laid out afresh for finger use: larger selects, functions as icons with small captions, each at least three centimetres tall, in four groups by task: the text and its chapters, the run (pace and voice following), the picture (fullscreen and black), work on the text (edit, review, settings). Every button always stays in the same place. While following your voice the pace slider turns grey and shows the measured speaking pace; a tap on the number adopts it.

The selects in the header closed by themselves in fullscreen, because the bar slid away under your finger or fullscreen was requested again. As long as someone is using the bar, it now stays put.

The note “screen may go dark” is dropped while the device is plugged in, provided the browser reports it. Safari does not, so the note stays there as before.

In the key list the buttons showed in browser white and the actions at the bottom were stuck together; in the dark colour schemes the scrollbar was also black on black. The browser’s controls now follow the colour scheme, and the buttons look the same everywhere.

The home page has two new sections: On set, with foot pedals, pairing by QR code, fixes from the phone, the second prompter and the time left, and From the manuscript to sign-off, with import from Word and PDF, the editor, sign-off by your client and versions. Plus a section for the two free tools.

The features page is split into two groups, in front of the camera and at the desk, and has five new blocks: the editor, import from Word and PDF, sign-off with comments, editing live from the phone and a second prompter on set.

Language is now switched through a dropdown with full names instead of four abbreviations side by side. It sits in the same place at every screen width. The hint for visitors whose browser prefers another language is unchanged.

The pricing table now also names the three paid features that were missing: sign-off with comments, editing live from the phone and the second prompter. Import from Word and PDF has a row of its own, and it is included in both tiers.

Help has two new sections, getting scripts in and sign-off by your client, plus two cards on set about pairing by QR code and the location. The reading time calculator now sits in the footer of every page.

The example on the features page showed the finished display under as you type it, so WELCOME and TANJA, instead of the notation with a hash and an at sign.

In French and Spanish the header wrapped onto two lines on medium sized screens, because the labels are longer there. Up to 1280 pixels the main menu now sits in the drop-down menu; the wordmark, the language, sign in and the button stay visible.

After dragging with your thumb, the text on the phone briefly jumped back to the old spot before landing on the right one. The cause was the prompter’s earlier report still on its way. The phone now skips every report that does not yet know about its own jump. The same goes for chapters, start and switching text.

When the text stood still during voice following because it was waiting for your voice, the phone still ran on for a line and was pulled back right afterwards. The prompter now reports how fast it is actually moving at that moment, instead of the phone guessing from the pace slider.

Jumping to a chapter now puts its first spoken sentence on the reading line, not the chapter marker. Before, depending on spacing, the last line of the preceding paragraph stayed on the line.

Mirror, upside down, lock, black and voice following now show their new state immediately, instead of snapping back on release and only changing seconds later.

The settings can be scrolled when they no longer fit on screen, and the explanation for the location appears in a floating bubble instead of in the middle of the form.

The text has more room: the side rails are narrower, and the gap between them and the text is gone. The buttons on them are also dimmed and only become distinct when touched, so they do not pull your eye away from the text in front of the camera.

Speech recognition could stall silently: the microphone stayed on, the tab icon kept glowing, and the text still stopped advancing. Usually after a system notification, or when the window had briefly been in the background. The only cure was switching voice following off and on again. The prompter now notices this itself and restarts recognition without you doing anything.

The browser’s microphone prompt was invisible in fullscreen: it waited until fullscreen ended, and until then it looked as if “voice following” did nothing. The first time around, the prompter now leaves fullscreen itself, and your next tap brings it back.

The status line now says “following your voice” only while something is actually arriving. After eight silent seconds it goes back to “listening”.

New in the prompter settings: the location. Enter a name there and this prompter only listens to remotes with the same name. Two set-ups in the same project can then run side by side without starting, switching or stopping one another. Left empty, everything stays as before: every device in a project controls every other.

The QR code under “pair remote” carries the location with it. Whoever scans the code of a particular iPad controls that one and not the one next door, without choosing anything on the phone. Which location a phone is driving is shown at the top, next to the connection indicator.

The paired remote showed neither text nor chapters: the button created a link with the permission “remote only”, and that one is text-blind. It now takes the most modest link the project already has, and creates one only if there is none at all.

If you press “pair remote” in the prompter and have no matching share link yet, one is now created there and then, instead of a note telling you to create one first. It is called “created on set” and appears in the project’s list of share links like any other: if you do not want to keep it, delete it there.

If you stop mid-sentence while following your voice, the text now coasts to a halt instead of running on for another line. How far it carries on depends on the rhythm in which speech recognition delivers its results, no longer on a fixed time.

When you carry on speaking, the text no longer slides back. If it is a little ahead of you, it waits until you have caught up instead of showing you the same line a second time.

The slider is now called “width” and means the whole column: 100 per cent fills the space between the side rails, and down to 30 per cent the line gets shorter. Before, the width was counted in characters and therefore grew with the type, so “size” quietly moved the column as well.

“Size” is now called “text size” and changes nothing but the text size.

The dark fade at both edges now lies over the text only, no longer over the header row, where it used to sit on top of the wordmark and the hint.

If the project has no share link yet, “pair remote” now says so instead of showing an empty link with no code.

In the key list the scrollbar sat on the cross that clears a key. It now has a margin of its own.

“Who are you?” now only asks when the text actually has two speakers, and only once per device. Answering “nobody in particular” is remembered; before, the question came back at every change of role. It can be changed at any time under “I am” in the settings.

The question “another prompter is open in this project” stays answered. Whoever chose “I lead” keeps leading, even when the other device briefly disappears, for instance because its window is throttled in the background; that is exactly what brought the question back after every switch to full screen. And a reload no longer mistakes itself for the second prompter: the window keeps its identity across a reload.

Every page is now built the same way: a header with a way back, a title and the one action, below it cards with one topic each. Anything that cannot be undone asks in its own dialog that names the action, instead of a browser box saying “OK”.

Two areas instead of five menu items: Projects and Settings. All settings live in one place, grouped into Account and Personal; before, they were split between the main bar and the user menu. The start page is the project list, and there is no second page with the same list any more.

States show as badges with five tones, no longer as loose text in changing colours. Empty lists say what would be here and offer the action that gets it there. Invoices sit in a table with their state.

The operator area has a face of its own: a dark sidebar with the groups Overview, Customers, Money and System. The overview now shows what running the service is about: paying accounts, monthly value, revenue this year and last month, open and overdue amounts, new accounts, cancellations, and below that, what needs doing.

New pages Payments (paid invoices by month with sums, the page for bookkeeping) and Cancellations. The account list shows plan, since when, open amounts and last activity, with filters; the account shows total revenue, open sum, people, projects and every invoice with its actions.

The operator area now keeps a log: who set a plan, booked or cancelled an invoice, locked an account, set a password, changed a price or a setting is listed under Log with time, person and target. Passwords and similar secrets never appear there, and after a year the nightly run clears out old entries.

The editor shows the review comments again, including resolving and reopening them. The list had been lost while rebuilding the interface, and opening a script got stuck.

The suggestion list after “@” no longer shows half typed names. And anyone who picked a suggestion with the arrow keys or the mouse can now accept it with Enter; with nothing picked, Enter still starts a new paragraph.

The writing area starts tall enough to work in. With an empty script it used to be exactly one line.

Clicking a chapter in the outline puts the heading at the top edge. It used to land on the bottom line, leaving the chapter itself out of sight.

A “[Pause 3]” now counts towards the running time. Chapters, total length, the prompter’s remaining time and the comparison with the target all include the held seconds, and the outline says separately how much of it is pause. A chapter with ten seconds of silence used to promise ten seconds too little.

A pause may now be written the same way everywhere: “[Pause 2]”, “[Pausa 2]”, “[Break 2]”, with or without a space and with a trailing s. The prompter used to hold only on the German spelling, while the reading time calculator on the website counted all three.

The reading time calculator counts the text after a speaker name. With “@Anna: Good morning” the whole line used to be dropped, and the calculator gave a shorter time than the prompter.

When uploading a file, the limits showed as placeholders instead of numbers: “At most :size, :pages pages and :characters characters.” The values are there now. The same hit the column choice, the number of pages read and the open notes in the editor.

An account can be given full access free of charge. It then never receives an invoice and does not expire. The billing page says so in a sentence instead of offering a renewal that does not exist.

There is now a full teleprompter without signing up: paste your text, set the pace, go. With speakers, chapters, directions, voice following, mirroring, black screen and reading time.

Your text stays in this device's browser. It is not uploaded and not stored, and a button deletes it again. That Safari clears storage like this after about a week is on the page, not in the small print.

If you want to take the text into an account, one button does it. It travels in a URL fragment, and browsers never send those to a server: this route too keeps the text on your device.

Plus a reading time calculator: paste the text, pick a speaking pace, see duration and word count. Counting works exactly as in the prompter, pauses from the directions are added, speaker names and chapters are not.

What the free prompter cannot do is listed openly next to it, with the reason: the remote, the emergency mode, several texts, working together and your own AI all hang on an account.

Not a single mail was going out, and it failed silently: address confirmation, invitations, the collected review mail and the billing reminders all got stuck. The cause was a missing line in the mail settings. They arrive now.

Two devices show the same text at the same spot: one for you, one for your guest. Type size, reading line, mirroring, colour scheme and the highlighted speaker colour stay per device. On one, Ben is bright and Anna dimmed; on the other the other way round, both on the same line.

The warning “two prompters are reporting” becomes a question with two answers: “I lead” or “I am the second prompter”. Whoever does not answer now does nothing at all. That solves the second window somebody left open: until now it carried out every key press a second time.

The remote now says what is going on: one prompter (then it keeps quiet), two without coupling (with what to do about it), or two coupled, by name. Typeface and colour scheme can be set separately for each of the two.

When you couple, we ask once who is standing at this device, with the speaker list of the running text. The name then shows in the remote so you can tell the two devices apart.

Coupling needs a network, and it says so. Without one, each prompter keeps running on its own, just as before.

A typo turns up on set: press and hold the paragraph, correct it, done. The iPad shows the right sentence two seconds later and stays exactly where it was. The run does not stop.

You edit the paragraph, not the whole text. A text box with eight thousand characters, half hidden by the on screen keyboard, is unusable on set; a paragraph is three lines.

If somebody else saved the same text meanwhile, we merge instead of asking, as long as the changes sit in different places. Only when that very paragraph was rewritten do we ask, with two buttons: fetch again, or overwrite anyway.

While someone is writing on the phone, the edit button on the iPad stays locked, and the other way round the remote holds still while the iPad is being edited. The version list afterwards says that this state came from the phone on set.

You can now upload Word, LibreOffice, PDF and plain text. Headings become chapters, speaker names become speakers, and an italic run becomes a direction. Before you take it over you see the document on the left and the result on the right, side by side.

Two column templates, directions on the left and the spoken text on the right, used to come out as gibberish. Now we ask which column is the text and take only that one.

The uploaded file stays with us only as long as the reading takes. After that it is gone, and only the text remains.

A share link can now allow comments. Whoever has it reads the text the way the prompter will show it, marks a spot and writes about it. No account, no login, just a name.

A comment finds its place again even when the text has changed since. Where a paragraph was rewritten from scratch that no longer works: those comments stay on with their quote instead of quietly vanishing.

A round of review produces ten comments in five minutes. You get one collected mail for them, not ten, and you can switch it off in your profile.

In the editor you see the open comments on the text, you can tick them off, and you see who wrote them.

You now write speakers with an at sign and chapters with a hash: @Tanja and #Opening. That fixes three faults at once. Names with more than one part work now (@Anna Miller), they no longer end up silently as a chapter, and @Tanja: Good morning on one line works. The model is Fountain, the standard for screenplays as plain text.

A line entirely in capitals is now ordinary text and gets read out. Before it silently became a chapter, and that was exactly the fault: capitals are a perfectly normal way to mark emphasis on a prompter. The editor points it out when a line looks like a chapter; in the prompter nothing warns, that is where you shoot.

The existing texts have been moved over without anything changing: same chapters, same speakers, same reading time. The previous state sits next to it as a version. Where a text was left behind, the editor shows a note with a button, and the prompter on set still accepts an old text without complaint for a while.

The editor now colours what it sees: speakers, chapters, directions and pauses. A forgotten closing bracket is visible at once, because everything behind it carries on dimmed. On the right sits the outline with chapters, reading time and speakers, and a click jumps there. At @ and # it suggests what the project already has; Enter stays Enter, you accept with Tab.

Seven hints that pay for themselves: target time missed, unclosed bracket, very long paragraph, chapter without text, paragraph without a speaker, mistyped speaker name, and a pause that gets silently cut. Hints are information, not accusation, and they never hold up saving.

Clean-up tidies what comes along when you copy from Word or a PDF: invisible characters, ligatures, hyphens at line ends, bullet leftovers. None of it happens silently. You tick, you see the preview with a count per rule, and you apply; one single Ctrl+Z takes the whole run back. The content inside square brackets and your annotation lines are never touched, and German quotation marks stay as they are.

Text pasted from Word and Google Docs is recognised and shown in a preview: „Recognised from Word: 3 speakers, 4 chapters, 7 directions.“ Apply, paste as plain text or cancel, you decide. Pasting never rebuilds silently. If only plain text is on the clipboard nothing happens at all, and a quote from your own text never runs through a dialog.

Signing in and registering did not work in the browser: both forms called an address the frontend did not know, and the button silently did nothing. Fixed, and a guard now watches over it.

Foot pedals, presenters and Bluetooth clickers now drive the prompter. Which key does what is in the settings under Key list, and every line can be changed: press Learn key, press your key, done. At the top you always see what arrived last. If you do not know what your pedal sends, you find out there in ten seconds.

Once you had tapped the script picker in the header, your keyboard and pedal were dead: the browser scrolled inside the select instead. That is gone. The prompter now listens first, and every control gives the focus back afterwards. Only while writing does it stay out of the way.

Three fonts, five colour schemes, three contrast levels, plus line spacing, letter spacing and column width. Amber on black for long shooting days, dark on light for outdoors in the sun, cream on blue-grey because pure white on pure black flickers. All of it is free, in every plan, and all of it can be set from the phone as well.

The prompter loads a third less: the base font is now one file per character set instead of twelve and can still do every weight. Lexend and OpenDyslexic are fetched only when someone picks them, and then move into the device's store. The settings tell you whether the chosen font is already there without a network.

With a paid plan the account remembers your keys and your appearance. A new iPad on the shooting day then has them straight away. Setting them is open to everyone, free access included: on the device they apply either way, even offline.

French and Spanish are translated and now available to choose, in the application, in the prompter and on the public site. That completes all four planned languages. The addresses were already there and do not change: /fr/tarifs with German text becomes /fr/tarifs with French.

The French and Spanish pages are now allowed into the search index and appear in the sitemap and in hreflang. Except for the legal texts: those exist in German only, and wherever the German version appears the page says so and keeps itself out of the index.

Two commands help with translating: prompterpal:lang-missing shows which lines are missing in a language, prompterpal:lang-unused finds keys nobody uses in the code any more. Both only look and change nothing.

www.prompterpal.de now has content: home, features, pricing, who it is for and help, plus the legal texts and this page. All of it in German and English, with its own address per language: /preise and /en/pricing. French and Spanish already have their addresses but carry German text until they are translated, and stay out of the search index until then.

The help page explains the text format with speakers, chapters, stage directions and [Pause 2], the moves that matter on set, and the keys the prompter listens to. Plus a section of its own on working without a network, which also says what does not work and why: following along runs at Apple and Google, and the remote needs our server.

PrompterPal has a mark: the “P” in a circle, as a favicon, as the home screen icon and in the preview image when somebody shares a link. The wordmark lives in one place in the code, so a later logo arrives everywhere at once.

On the public site the address now decides the language, and nothing else: /preise is German, /en/pricing is English, whatever the browser says. Anyone who prefers a different language gets a pointer to it at the top. Before this, the same address could serve two languages depending on the browser, and search engines saw one or the other at random.

You can now store your own AI key, under “Your own AI” in your menu. Three providers to pick from: OpenAI, Anthropic, and any other address that speaks OpenAI's format, which covers DeepSeek, Mistral, Groq and whatever comes next. “Check connection” fetches the model list once, so you can see the key works and choose a model right away. We provide no AI ourselves and bill nothing: what your provider charges is between you and them.

The editor now has a “Revise” button. Six ready-made instructions plus a free field: more relaxed, shorter to the target time, translate, add stage directions, make it easier to speak, spelling. The suggestion appears while it is being written, then stands next to your text with the differences marked. You take it over paragraph by paragraph, and only what you tick. What you take over is recorded as a version together with the instruction you used.

Nothing of a revision stays with us: no request, no answer, no log. Only what you explicitly take over becomes a version of your text. Your key is stored encrypted, belongs to you personally rather than to your account, and is never shown again after you save it.

You can now switch to Pal, by card, SEPA direct debit or invoice. Pay by invoice and your access opens right away; the invoice arrives as a PDF by email and is due in 14 days. All your invoices are under “Billing” to download.

The limits of the free plan now apply, and they apply gently: one project and five scripts are free. If you have more and later end up on Free again, you lose nothing. Anything beyond the limit stays readable and you can still run it in front of the camera; editing works again once you are back within the limit. Nothing is ever deleted.

You can end your contract at any time using the cancellation button on our site, without signing in. You get a confirmation in writing right away plus a link; only clicking it carries out the cancellation, so nobody can cancel on your behalf. It takes effect at the end of the period you have paid for.

Imprint, privacy policy, terms and the withdrawal policy including the model withdrawal form are now on the site.

The operator area has arrived: view accounts and users, book bank transfers as paid, set prices and limits, maintain the legal texts, lock accounts and rescue them in an emergency.

PrompterPal now has accounts. Register and your account comes with it; projects, scripts and later the billing hang from it. The customer is the account, not the person: one person can work in several accounts and still have one of their own. Signing in comes with a confirmation mail, forgotten passwords, two-factor authentication with recovery codes, passkeys, and the list of devices you are signed in from.

You can invite people into your account by email, with one of three roles: read and operate, edit as well, or manage the account. What gets invited is an address, not a user: anyone without an account creates one when accepting. Rights apply per account, not per project.

Projects bundle the scripts that belong together in front of the camera. Every script has a language and a target time, so an overlong script shows on set and not in the edit. Every save keeps a version you can bring back.

The prompter and the remote control from wedoro now run here, with everything they could do: following along by microphone, chapters, reading times, mirroring, black screen, and the emergency mode that keeps running without a network. The scripts come from the database instead of files; the format has not changed.

The person in front of the camera often has no account and wants none. Hence share links per project, with an expiry and a graded permission: read and run, remote control only without seeing the text, or edit as well. The phone picks up the link to the remote as a QR code from the iPad's screen.

The application speaks German and English. French and Spanish are prepared and will follow once they are translated.