Skip to content
E-commerceFind, track and return orders.BankingSpending, cards and transfers.TravelConnections and rebooking.TelecomBills and roaming passes.InsuranceClaims from one sentence.
AnswersTheir own data, shown as your app’s cards instead of a paragraph.ActionsTasks done through your API, confirmed with a tap or undone with one.VoiceThey say what they need and the app does it.MemoryRemembers what each person prefers and uses it the next time they ask.Your brandLooks like your app on iOS, Android and the web.Setup by your agentYour coding agent adds it and fills in the console.EngageMessages people asked for, written for you and sent with one tap.AnalyticsWhat people did, what they asked, and what got solved.Human supportYour team answers inside the app, with the conversation in hand.ImproveAnswers ready for what it missed, and requests counted by who asked.Why RendelA chatbot talks about your app.How it worksThe six steps of a turn.SecurityWhat nothing happens without.CompareNext to the alternatives.
DocsEverything, in eighty pages.FlutterThe package, on iOS and Android.WebOne script tag.React NativeSoonPlanned.SwiftSoonPlanned.KotlinSoonPlanned.
BlogWhat we are working out.GlossaryThe words, one sentence each.
Voice

Ask out loud. Get it done in the app.

Rendel is the AI assistant that works inside your mobile app and website. Your customers hold the microphone and say what they need, then get the answer as your app’s own cards, with its opening read aloud once you turn voice on.

Try the consoleBook a Demo

Halden’s console, an invented store with Rendel running. Yours for two hours, no sign-up.

Asked aloud, answered in the app

Halden, an invented store. Mira holds the microphone, asks, and hears the answer’s opening.

Read aloud“Your Trail Runner 2 is out for delivery and should arrive today, by 6 pm.”Hear the voices
The problem

Saying it is quicker. Adding voice usually isn’t.

On a phone, typing the question is often the slowest part of getting help. Teams that want voice usually take on a second assistant to do it, with its own prompt, its own tools, its own bill, and a reply with nothing to show on screen.

Typing slows the question down

In a 2016 Stanford study, people entered short English phrases on a phone about three times faster by speaking than by typing, and a slow question is an easy one to give up on.

Read the Stanford study

A voice tool is a second assistant

Voice-agent platforms need their own prompt, tools wired to your backend, speech and voice providers, and tokens from your server. Then two assistants have to give the same answers.

A spoken reply has nothing to show

A voice can say an order is on its way but cannot show the tracker, and a change still needs a way to be approved. Whatever appears on screen is yours to build.

A brand said wrong, to every customer

When a voice says your brand or your products wrong, every customer hears it, and that can be reason enough to leave voice off.

How it works

Ask out loud, hear the answer, confirm or undo with a tap.

Voice opens the assistant your app already has: people talk instead of typing, and hear the answer instead of reading it.

  1. Your team turns it on

    One switch under Assistant › Voice, with what it would add a month shown before it goes on. Pick a voice for each language, then publish.

  2. They hold the microphone and ask

    Their words are written large as they speak, in their own language, and letting go sends them.

  3. They get the same answer as typing

    An answer from their own account, drawn as your app’s cards, with its opening read aloud as it is written.

  4. Changes wait for their tap, or come with an Undo

    When the answer needs a yes, its question is read aloud too, and the change waits for their tap on Confirm, or runs with an Undo where your team allows it.

Speak

Ask in a sentence instead of typing one.

With a current SDK, the microphone is on by default in the message box, so your customers can talk from the first day.

Hold, say it, let go

People hold the microphone and talk. Their words appear large as they speak, and letting go sends them as the question.

  • On by default
  • Works by keyboard and screen reader
  1. A tap sends nothing
  2. Hold to speak
  3. Let go to send

They speak their own language

The microphone listens in your app’s language, or the phone’s when the app sets none, and the answer comes back in the language it heard.

A refused microphone says where to allow it

If someone said no to the microphone, the assistant says where to switch it on. On iPhone it offers Open Settings, and voice mode opens only once the microphone does.

Spoken or typed, the same answer

A spoken question takes the same path as a typed one: answers from the person’s own account, the same cards and forms, the same tap to confirm or undo. There is no second assistant to keep in step.

The microphone in your app’s look

Your team picks its glyph, Microphone or Waveform, and publishes it with the rest of your look. It stays in the message box.

Hear

Hear the answer without reading it.

Off until your team turns it on. Then people hear each answer’s opening in a natural voice, and keep looking at the cards.

Every answer’s opening, heard without a tap

Once voice is on, the opening of each answer is read aloud in a natural voice as soon as it is written, whether the question was typed, tapped or spoken. Nobody has to press Listen.

  • Off until you turn it on
Read aloud“Returns are free for 30 days after delivery, on anything unworn with its tags.”Read aloud“Your Trail Runner 2 is out for delivery and should arrive today, by 6 pm.”Read aloud“Size 42 is in stock. Confirm and the Court Low goes straight into your bag.”

The voice starts before the cards finish

The opening goes to the voice the moment it is written, so people hear it while the cards under it are still arriving.

It reads what matters and leaves the cards to look at

People hear the answer’s opening, at most two sentences. When it asks them to confirm, the card’s question is read too, and an answer that is only a result reads the result. The cards stay on screen.

Each person keeps their own Voice replies switch

A switch in the assistant’s header turns reading off on that person’s device. It starts on for everyone and is remembered between launches.

A screen reader hears: “Answers play by themselves”

It speaks only while they’re there, and never late

Reading stops when the app is put away, the tab is hidden, a new question goes out or the microphone opens, and it never starts in the background. A voice that would start late stays quiet: the answer is already on screen.

  1. App put away: it stops

  2. Microphone pressed: it stops

  3. Too late to start: it stays quiet

Heard with the iPhone on silent

In your iPhone app the answer plays with the ring switch on silent, and music or a podcast dips under it, then comes back. On the web, Safari is asked for a playback session where it has one.

  • Mobile SDK

Still heard where streaming isn’t

A browser that cannot play a stream gets the spoken opening as one audio file instead. If no voice can be made on our side, the phone’s or browser’s own voice reads it.

Voices

A voice your customers will like hearing.

Choose from 12 named voices, listen to each one first, and hear your own sentence before your users do.

12 voices, free to listen to

6 voices native in Turkish and 6 native in English, women’s and men’s. Each language gets its own, and every sample plays free before you choose.

English

HopeCurrentFemale · natural, calm
LaurenFemale · friendly, caring
KatieFemale · British, clear
BrianMale · deep, reassuring
JoshMale · warm, steady
DarianMale · warm, grounded

Türkçe

KateCurrentFemale · friendly, warm
EdaFemale · smooth, clear
BetülFemale · clear, attentive
OzanMale · rich, warm, calm
YunusMale · young, lively
AdamMale · calm, informative

Every sample reads:

English“Your order is on its way and arrives tomorrow between ten and noon. Shall I send you a reminder in the morning?”

Türkçe“Siparişiniz yola çıktı, yarın saat on ile on iki arasında kapınızda olur. Sabah size bir hatırlatma göndermemi ister misiniz?”

One voice for both languages, if you prefer

5 voices read Turkish and English alike, each keeping its own accent in the other language: Kate, Eda, Yunus, Adam, and Darian. One switch sets both at once.

Same voice in both languages

Both languages

KateCurrentFemale · friendly, warm · native Turkish
EdaFemale · smooth, clear · native Turkish
YunusMale · young, lively · native Turkish
AdamMale · calm, informative · native Turkish
DarianMale · warm, grounded · native English

Hear your own sentence first

Owners and admins write a line, a product name or a price, and hear it in the chosen voice with your pronunciations. Up to 200 characters and 30 a day per app, each counted like an answer read aloud.

  • Owners and admins

A retired voice never leaves answers silent

If a voice is withdrawn, the answers it read move to the nearest one, and the Voice page tells you which, so you can pick another.

Eddie is no longer offered, so the answers it read are read in Josh now, the nearest voice. Pick another below if you like.

JoshCurrentMale · warm, steady

Other languages are still read aloud

Native voices cover Turkish and English. An answer in another language is read by your English voice, or the one you chose before, and the voice is told which language it is reading.

Said right

Your names and numbers, said the way people say them.

Your brand and product names are said the way your team writes them down, and prices, dates and codes are written out before the voice reads them.

Your brand and product names, said your way

Write a word as it appears in answers and how it should sound, for every language or just one. Up to 500 words, and a Turkish ending after an apostrophe is kept.

Pronunciation

Make sure your brand and product names are said right: write a word as it appears in answers, and how it should sound.

WrittenSay it asLanguage
Trail Runnertreyl ranırTurkish
Karaköykah-rah-KOYEnglish
HaldenHAL-denAll languages
LoopbackluupbekTurkish

4 of 500 words.

The Turkish voice now hears: treyl ranır iki kargoda, yarın on sekize kadar kapınızda.

and keeps the ending: Halden’inHAL-denin

Add a whole catalogue, and see each row first

Paste a list or upload a CSV, TXT or Excel file from a template. Every row is checked before anything is saved, and a word you already have is replaced unless you choose to keep yours.

Add many words

  • Trail Runner → treyl ranırAlready there
  • Aero Knit → eyro nitNew
  • Court Low → kort louNew
  • Halden → HAL-dunYours stays: "HAL-den"Said differently now
Add 2 words6 of 500 words after this

Turkish numbers, said the way people say them

Prices, dates, hours, ordinals, percentages, units, codes, order numbers and phone numbers are written out in words before the Turkish voice reads them. Anything that could be read two ways stays as written.

Turkish numbers, said the way people say them
WrittenHeard
Court Low $118Court Low yüz on sekiz dolar
1.299,90 TLbin iki yüz doksan dokuz lira doksan kuruş
24.10.2026yirmi dört Ekim iki bin yirmi altı
14:30on dört otuz
3. katüçüncü kat
%20yüzde yirmi
750 mlyedi yüz elli mililitre
Sipariş #48213Sipariş dört sekiz iki bir üç
0850 123 45 67sıfır sekiz yüz elli yüz yirmi üç kırk beş altmış yedi

Hours, codes, addresses and dates, read in full

English hours are said on the 12-hour clock and codes character by character. In every language a street abbreviation is said in full and emoji are dropped, and short weekdays and months are said in full in English, Turkish, German, French and Spanish.

Hours, codes, addresses and dates, read in full
WrittenHeard
Open 9:00–18:00Open 9 a.m. to 6 p.m.
Order H-20418Order H 2 0 4 1 8
15 Riley St15 Riley Street
2026-11-1414 November 2026
Mon, Oct 12Monday, October 12
Pzt 12 EkiPazartesi 12 Ekim
Your return is ready ✅Your return is ready
Control and cost

On when you say, at a cost you see first.

Voice is one switch for the whole app. The console shows what it would add before it goes on, and it reaches your users when you publish.

One switch for the whole app, no app update

Voice stays off until your team turns it on. Once you publish, every device follows, and turning it off silences all of them at once.

See what it adds a month before you turn it on

Turning voice on shows what it would add at your app’s own traffic, in messages and dollars, and asks before it goes on. Once it is on, the page shows what it actually added in the last 30 days.

Voice answersYour users hear each answer as it arrives, without having to read it.

At your current traffic, turning it on adds about 790–1,310 messages a month (≈$32–$52). How much depends on how many of your users keep Voice replies on.

Halden’s invented traffic: 6,000 answers in 30 days on the Growth plan, worked out by the console’s own estimate.

Try it in your draft, then publish

Voice changes wait in your assistant’s draft with Experience and Appearance. An app build on a test key hears them at once, and your users hear them once you publish.

Charged at what the voice costs, and never twice

Each answer read aloud adds a quarter message for each started $0.005 of voice, apart from the answer itself. Speaking instead of typing adds nothing, test keys use none, and an answer fetched twice is counted once.

  • Every plan, Pilot too
  • Not billed by the minute
  • Voice up to $0.005: ¼ message
  • Up to $0.010: ½ message
  • Speaking instead of typing: No extra
  • On a test key: Free

Counts towardyour plan’s messagesyour monthly budgetthe app’s daily spend cap

Every voice change on the record

Turning voice on or off, choosing a voice and editing words are each written to the audit log, with what changed.

Audit log

  • Voice settings changedSam Carter · just now
  • Look publishedPriya Nair · 12d ago
  • Knowledge source addedPriya Nair · 14d ago
Try it on Halden

Speak to Halden, then choose how it sounds.

Halden is an invented shoe store with Rendel running for five weeks. Its console is yours for two hours, with no sign-up. These are the steps to take in the demo.

Try it in the demo console

Your own copy of Halden for two hours, no sign-up. Speak to the Test phone, then listen to the voices under Assistant › Voice.

Try the console

Voice in the assistant you have, or a second one beside it.

WhatWith RendelWith a separate voice agent
A spoken question getsYour app’s own cards and forms, from the person’s own account, with the opening read aloudA spoken reply. Anything on screen is yours to build
Doing somethingThe same actions as typing, through your API as the signed-in user, after a tap on Confirm, or with an Undo where you allow itYou connect each tool to your backend and design how people approve
Setting it upA switch under Assistant › Voice on the assistant you already run, then a voice and your wordsA second agent: its prompt, tools, speech and voice providers, and tokens from your server
How it is chargedPer answer read aloud, in your plan’s messages, with the monthly estimate shown before you turn it on. Speaking costs nothing extraUsually per connected minute, plus each provider’s own rate
How people talkInside your app: hold the microphone, speak, let go. No wake word, no talking over the voiceOften a live two-way call that people can interrupt, on a phone line or a widget
Voices and languages12 voices, native in Turkish or English; other languages read in your English voiceOften a large voice library, voice cloning and many languages
What your team seesSpoken questions are kept and reviewed with the typed ones, in the same conversations, not marked apartCall transcripts, in the voice tool
The customer’s audioNever reaches Rendel: the phone’s or browser’s recogniser turns it into wordsStreams to the voice vendor to be understood

If you need phone calls answered, a voice agent built for calls is the right tool. Rendel works inside your app and does not take calls.

Control and privacy

Your team decides, and each person can turn it off.

Who can change voice, what each customer controls, and what happens to what they say.

Who can change voice

RoleTurn voice on, pick voices, edit wordsHear your own sentenceThe Voice page
OwnerYesYesYes
AdminYesYesYes
DeveloperYesNoYes
Analyst, Conversation reviewer, and Customer supportNoNoRead only

Read-only roles see: “Your role can see these settings but not change them.”

Rendel never receives their voice

The microphone opens only when the person presses it, with no wake word. The phone’s or browser’s own speech recognition turns speech into words, and only the words reach Rendel; that recogniser is Apple’s, Google’s or the browser maker’s, and may send the audio to them. Rendel neither receives nor stores the audio, and a spoken “yes” never confirms a change.

What they say is kept like what they type

A spoken question is stored like a typed message, for the conversation retention your team sets. Emails, phone numbers and card numbers that users type are masked before the model sees them.

Only the opening goes to the voice

To be read aloud, an answer’s opening, at most about 400 characters, goes to the text-to-speech provider, ElevenLabs or OpenAI, in the United States. It can include an order number or an address. The audio is streamed to the device and not stored.

Where it is kept

Rendel’s database is in Frankfurt (AWS eu-central-1); the voice providers are in the US. The words travel sealed, never in a web address in the clear, and the usage record keeps seconds and characters, not the words. A phone that falls back to a whole audio file keeps only that file, until the next answer.

Your console can’t silence one person

Only that person can, with Voice replies in the assistant’s header, which their device remembers, or your team can for the whole app.

Your developers have the last word

In code they can remove the microphone, keep one app or site silent, or use their own speech recogniser. Voice plays only where both the code and the console allow it.

SubprocessorsPrivacy policyData processing agreement

Turn it on

What it takes to turn on

The microphone already works with a current SDK. Reading answers aloud takes one switch and a voice.

It comes with a current SDK

Hold to speak needs Mobile SDK 0.60.0 and Web SDK 0.39.0 or later, and answers are read the way this page shows from Mobile SDK 0.59.9 and Web SDK 0.38.9. The newest are Mobile SDK 0.82.0 and Web SDK 0.57.0. Android apps on the newest Flutter need Mobile SDK 0.69.1 or later.

On iOS, two permission lines

Add NSMicrophoneUsageDescription and NSSpeechRecognitionUsageDescription to Info.plist, or the microphone is simply hidden. Rebuild the app fully once, since the voice player is native code.

On Android, nothing to add

The SDK’s plugin declares the microphone permission.

On the web, https

The microphone needs https. Inside an iframe, add allow="microphone; camera". Browsers without speech recognition, Firefox among them, show no microphone.

Every plan has voice, Pilot included

Nothing to upgrade: answers read aloud count toward your plan’s messages like the answers themselves, and stay off until you turn them on.

In the console

  1. Open Assistant › Voice and turn on Voice answers. The dialog shows what it would add a month; press Turn on.
  2. Pick a voice for each language, after listening to the samples.
  3. Add your brand and product names under Pronunciation, if the voice says any of them wrong.
  4. Publish. An app build on a test key hears it before you do.
  5. Optional: set a monthly limit under Settings › Usage limits, watch the voice line under Plan & usage, and choose the microphone’s glyph in Appearance.

In the docs

Answers read aloud: settings and costThe consoleWeb SDK: permissionsWeb SDK: optionsMobile SDK: RendelConfigPlatform permissions

What teams ask before they start.

  • Yes. They hold the microphone in the message box and speak, their words appear as they talk, and letting go sends the question. They get an answer from their own account on your app’s cards, and a tap to confirm or undo any change.

  • Yes, once your team turns voice on under Assistant › Voice. Then the opening of every answer is read aloud in a natural voice as it is written, however the question was asked. Each person can turn that off for themselves with the Voice replies switch in the assistant’s header.

  • No. People hear the opening, at most two sentences. When an answer asks them to confirm, the card’s question is read after it, and an answer that is only a result reads the result’s words. The cards themselves stay on screen to look at.

  • The microphone listens in your app’s language, or the device’s when the app sets none, so people can speak any language their phone or browser recognises. Answers are read in {voices} voices native in Turkish or English, and an answer in another language is read by your English voice.

  • You decide how they are said. Under Pronunciation, write a word as it appears and how it should sound, for every language or one, up to {lexiconMax} words, or add many from a list or a CSV, TXT or Excel file. Prices, dates, hours and codes are written out before the voice reads them.

  • Yes, the same actions they could ask for by typing, through your app’s API. A change to their account shows a confirmation card with its question read aloud or, where your team allows it for something easy to take back, runs with an Undo they can tap. Saying yes out loud never confirms anything.

  • No. The phone’s or browser’s own speech recognition turns speech into words, and only the words reach Rendel, where they are handled like a typed message. That recogniser belongs to Apple, Google or the browser maker and may send the audio to them. Rendel never receives or stores it.

  • Speaking adds nothing: a spoken question costs what a typed one does. Each answer read aloud adds a quarter message for each started {quarterUsd} of voice, in your plan’s messages and budget, and test keys use none. Before you turn voice on, the console shows what it would add a month at your traffic. It is not billed by the minute.

  • The microphone is: with a current SDK it is on by default in the message box. Reading answers aloud is off until someone on your team turns it on, and like the rest of your assistant it reaches your users when you publish.

  • Yes. Your developers can turn reading aloud off in code for one app or one site, and remove the microphone the same way, while the console’s switch keeps voice on everywhere else.

  • A current Mobile or Web SDK. On iOS, two permission lines for the microphone and speech recognition, or the microphone is simply hidden; Android needs nothing more. On the web: https, microphone permission inside an iframe, and a browser with speech recognition, which Firefox does not have.

  • No. People hold the microphone, speak and let go, inside your app. There is no wake word, and pressing the microphone while it reads stops the voice. If you need phone calls answered, a voice agent built for calls is the right tool.

Related

Put it in your app this week.

Start free on your own, or tell us what your customers do most and we’ll show you that exact workflow, working, in a tailored demo.

  • PilotFree400 messages a month
  • Starter$199 / month4,000 messages a month
  • Growth$799 / month20,000 messages a month
Start with the PilotBook a Demo
Every plan, and the estimatorOr try a live store’s console first, yours for two hours

Prefer email?

sales@rendel.ai

Prefer to write first?

Tell us about your app and we’ll show the right workflow in a tailored demo.

Help us tailor the demoOptional

We use this to get in touch about the demo and, if a partner introduced you, to credit them. We never sell it. See the privacy policy.