All industries

Offline voice prompts for apps and devices

A device that speaks needs hundreds of small sentences: a confirmation, an error, a setup step, a warning before something is erased. This pack is those sentences, already spoken and checked, so the audio ships inside your product and plays with no network call. The agent on top handles the support texts around it.

Questions callers bring to apps and devices

  • How do I set this up?
  • Is there an update for my device?
  • Will a factory reset erase my data?
  • My device says there is an error, what now?
  • Is it still under warranty?
  • Can I get a replacement?

What the agent does for apps and devices

These are ready-made workflows that ship with the agent for apps and devices. They start switched off, you read and approve each one, and nothing is sent in practice mode.

  • Points to your setup guide and offers a person, without walking anyone through steps you did not write. Ready-made workflow: Offers setup help.
  • Texts customers that an update is out. Ready-made workflow: Tells customers an update is available.
  • Passes replacement requests to a person and never promises one by text. Ready-made workflow: Passes replacement requests to a person.
  • Warns that a reset can erase data before the customer presses the button. Ready-made workflow: Warns before a reset.

What it will not say, and when it hands over

From the agent's built-in rules for this kind of business and its ready-made workflows. A person is reached with the keypad (press 0) or by the hand-off workflows.

  • Never: promise a replacement, repair or refund.
  • Never: ask for or repeat a password, key or code.
  • Never: give a fix time you did not approve.

What it connects to

  • GitHub or GitLab webhooks (setup guides)
  • PagerDuty incidents (setup guide)
  • Zendesk, Freshdesk, Help Scout or Jira Service Management through webhooks and the http action (named in our connector docs, no dedicated guide, not verified)
  • Any other system that can send a webhook, through the universal connector (a short mapping in agent.json)

GitHub, GitLab and PagerDuty have dashboard setup guides, which suit a product team. A helpdesk that can send a webhook works through the universal connector. We call a guide a setup guide, not a verification: test each connection in practice mode with your own accounts first.

What it costs against a monthly AI receptionist

Across all industries the agent is $214 to $399 once per business location, depending on the number of phrases (audio alone is $14 to $199); the pack table on this page shows the exact price for this industry. The first 12 months of support and upgrades are included. After that an optional $99 a year renews support and upgrades; what you downloaded keeps working for ever either way. For comparison, hosted monthly plans read from their own pricing pages on November 22, 2026: Goodcall Starter is $79 a month ($948 a year), My AI Front Desk Business-in-a-Box is $99 a month ($1,188 a year), Smith.ai Starter is $300 a month ($3,600 a year). Those plans cap calls, minutes or customers and are billed every month. Speakvora has no licence fee per call, text or minute, but your phone provider bills your calls and texts, and an AI key, if you add one, is billed by your AI provider. See how much an AI receptionist costs and the pay-once alternatives.

This is not a hosted service and not a live answering service. The agent runs on your own computer, which has to be on and online for it to answer. If you want nothing on your own machine, or a person to answer, a hosted plan is the better choice.

Related trades

voice prompts for self-service kiosks · voice support for smart home devices · ai phone support for it help desks

What an AI receptionist costs · Pay-once alternatives · Every agent feature · AI front desk by industry · All industries

Example lines

  • “Thanks for calling, what can I help you with today?”
  • “Hi there, how can I help you get sorted out?”
  • “Good to hear from you, what's going on with your app or device?”

Packs and prices

VoicePhrasesSizePrice (once): audio only / with the agent
ph_delia3,630125.5 MB$99 / $299Free sample · Buy
ph_alfie3,630120.2 MB$99 / $299Free sample · Buy
ph_easton3,630117.4 MB$99 / $299Free sample · Buy
ph_everly3,630126.4 MB$99 / $299Free sample · Buy
ph_juniper3,630125.1 MB$99 / $299Free sample · Buy
ph_pippa3,630125.8 MB$99 / $299Free sample · Buy
pk_jasper3,630113.7 MB$99 / $299Free sample · Buy

Also in other languages

Native-speaker review pending.

LanguageVoicePhrasesPrice (once): audio only / with the agent
archatterbox-ar500$24 / $224Free sample · Buy
Spanishef_dora702$24 / $224Free sample · Buy
Spanishem_alex547$14 / $214Free sample · Buy
fichatterbox-fi378$14 / $214Free sample · Buy
Frenchff_siwis617$14 / $214Free sample · Buy
Hindihf_alpha500$14 / $214Free sample · Buy
Hindihm_omega500$14 / $214Free sample · Buy
Italianif_sara500$14 / $214Free sample · Buy
Italianim_nicola500$14 / $214Free sample · Buy
Japanesejf_alpha534$14 / $214Free sample · Buy
Japanesejm_kumo525$14 / $214Free sample · Buy
Portuguesepf_dora617$14 / $214Free sample · Buy
Portuguesepm_alex500$14 / $214Free sample · Buy
Mandarin Chinesezf_001500$14 / $214Free sample · Buy
Mandarin Chinesezm_009500$14 / $214Free sample · Buy

When a line is not in your pack

A pack covers the common apps and devices lines, but a customer will sooner or later need one it does not have. With your API key the new line is voiced on demand:

  1. It arrives fast. In our November 21, 2026 test, an English catalog voice reached the first audible sound in 44 ms with a Phonon voice, from a client in AWS us-east-1 (our own measurement, using the Coval test method).
  2. You pay once for it. $3 per million characters, about $0.0002 for a 60-character line, paid when the line is voiced.
  3. It joins your library. The local agent (version 0.40 or later) saves the audio on your computer and maps it to a slot, so the next time it plays from disk, instantly, with no internet and no charge, and the same words are not voiced or charged twice. Saved lines keep playing if you later stop using the key; speaking new lines needs a working key.
  4. Your own AI can write it. If you connect an AI of your own, it can write the reply, and Speakvora voices it as soon as the text exists (how fast the words appear depends on that AI). Lines that hold a name, a number or a link are spoken but are not added to the library automatically, so customer details stay out of it.

The voices are AI-generated. Details and method: text to speech latency · grow your voice library.

What is in this agent

  • Library: approved, pre-recorded apps and devices lines, one payment per voice and location, no per-play fee, the first 12 months of support and upgrades included with the industry agent (afterwards everything you downloaded keeps working for ever; support and upgrades after that are an optional support plan at $99 a year per location), and you can request lines that are missing.
  • Agent: open-source, runs on your own machine with ready-made workflows for this industry. You configure it to handle the routine questions and events you choose, from your recordings: no per-turn or per-minute charge from us.
  • Offline: the recordings play with no connection; texts, calls and posts queue when the internet drops and go out in order when it returns. Anything that may or may not have arrived is held for your review, not sent again by itself.
  • API: the same voices through the speak API for any new sentence ($3 per million characters, $3 for your own voice).

More sample lines

These come from the free 25-sentence sample of the Apps and devices pack, so you can judge the writing before you buy. Each line was written for this situation and checked before it was added.

Alert

  • A certificate expires soon.

Turn

  • A faster route is available.

App onboarding and controls

  • A new update is available for your device.
  • A technical issue forced the application to stop.
  • Absolutely, I can walk you through disabling that right now.
  • Absolutely, you can adjust these settings anytime from your preferences menu.
  • Absolutely, your settings aren't locked and can be updated anytime.
  • Access to your camera is required for this function.
  • Acknowledged, you're comfortable navigating with the primary features.

Voice assistant general

  • A pharmacy is accessible from the parking area you're approaching.
  • A pharmacy is located approximately half a mile in the southeast direction.
  • A pharmacy with a full range of services is ready to assist you at that location.
  • Absolutely no problem, take care and have a wonderful day.
  • Absolutely. To set this reminder up, I'll need the day, time, and what it's for.
  • According to our system, here's what I'm seeing with your order.
  • After cancellation, you may resubscribe at any time in the future.
  • After this is done, you won't be charged again.

Gaming

  • Absolutely my pleasure, and thanks for being such a great player to work with.
  • After cancellation, you'll still be able to play as a standard member.
  • After deletion, you're welcome to start fresh with a brand new account whenever you'd like.

Workplace field service

  • Absolutely. All your information is documented and we're ready to assign a technician.
  • Absolutely. We have what we need and will begin coordinating the visit immediately.
  • Accepted methods are credit card, debit card, and check.
  • Accepted payment methods are credit card, debit card, and draft.
  • Acknowledged. The team will review your needs and confirm availability within one business day.

What the included agent does for apps and devices

Every pack includes the open-source pack agent. It runs on your own server, uses your own phone, Slack, Teams, calendar and email accounts, and only says the approved sentences. For this industry it ships with 85 ready-made workflows: 38 written for it and 47 that react to events in your app, for example:

  • Confirms a new booking: when your app sends an event, then a text, a calendar entry, a staff alert.
  • Confirms a changed booking: when your app sends an event, then a calendar change, a text.
  • Confirms a cancelled booking: when your app sends an event, then a calendar cancellation, a text.
  • Follows up kindly after a missed booking: when your app sends an event, then a text.
  • Helps customers change a booking by text: when a customer texts, then a reply, a staff alert.
  • Tells customers you are ready for them: when your app sends an event, then a text.
  • Passes questions that need a person to your team: when a customer texts, then a reply, a staff alert.
  • Answers parking and visiting questions: when a customer texts, then a reply.
  • Thanks customers and asks for a review: when your app sends an event, then a text.
  • Answers “what are your hours?” by text: when a customer texts, then a reply.
  • Answers “where are you?” by text: when a customer texts, then a reply.
  • Answers “where is my order?” by text: when a customer texts, then a reply, a staff alert.
  • Passes “I want a person” to your team: when a customer texts, then a reply, a staff alert.
  • Answers your phone with a menu: when a customer calls, then plays the phone menu.

It starts in practice mode, so nothing goes out until you switch it on, it always honours STOP replies, and the texts and calls need your own providers and an internet connection (when it drops, they wait in a queue and go out in order when it returns; anything it cannot confirm arrived is held for your review, not sent again by itself). How the agent works.

How people use a pack like this

  • Phone menus and voicemail greetings. Load the files into the phone system you already have.
  • Kiosks and devices. Ship the audio with the device so it speaks with no network call and no per-play fee.
  • App notifications. Play the right file when an event happens, such as an order or appointment update.
  • Anything not in the pack. Use the Speakvora API for new sentences and POST /v1/prepare to get your own lines ready ahead of time.

Before you buy

  • Format: M4A (AAC) audio, one file per line, with a manifest.json that maps each sentence to its file and a CSV.
  • Licence: commercial use in your own systems and your company group's systems; no resale or redistribution of the audio files. Read the licence first.
  • Voices: synthetic voices, not real people. Tell listeners where the law or good practice requires it.
  • Free sample and updates: every voice has a free 25-sentence sample, and updates and support are included for 12 months per business location (afterwards everything you downloaded keeps working for ever; the support plan is optional and never automatic).
  • Price: one time per voice, with no subscription and no meter.

Questions

Can I put these voice prompts inside my own app or device?

Yes. The licence allows commercial use inside your own apps, kiosks and devices, and your company group's systems. You may not resell or redistribute the audio files themselves.

Do the prompts play without an internet connection?

The audio files are stored with your product, so they play with no network call and no per-play charge. Only lines that are not in the pack need a working API key to be voiced.

What if my product needs a sentence the pack does not have?

Voice it through the API at $3 per million characters, with 100,000 characters free every month, and save it. A saved line is just another file.

What do I get?

A zip with one audio file per line, a manifest.json that maps each sentence to its file, a CSV and a licence. The largest apps and devices pack has 3,630 recorded lines; smaller packs have fewer, as the table shows. Every pack also has a free 25-phrase sample.

Can I use the audio offline and commercially?

Yes. You may ship the audio inside your own apps, kiosks, phone systems and devices. You may not resell or redistribute the files themselves.

How is it priced?

Paid once per business location. Audio only is from $99; with the local agent, dashboard and workflows it is from $299, with the first 12 months of support and upgrades included and an optional $99 a year after that. Compare human voice-over at about $80 per 100 words. What you download keeps working for ever.

Are the voices real people?

No. They are synthetic voices. Disclose that to listeners where the law or good practice requires it.

Can I get lines that are not in the pack?

Yes. With your API key a new line is voiced on demand: in our November 21, 2026 test the first audio reached the listener in about 44 ms for an English Phonon voice, and the local agent can save the line on your computer so it plays instantly and free from then on.

What does a new line cost, and do I pay again for it?

$3 per million characters, about $0.0002 for a 60-character line, paid once when the line is voiced. The agent keeps the audio on your computer, and the same words are not voiced or charged twice.

Do saved lines still play if I stop using my API key?

Yes. Saved lines are files on your computer. They play with no internet and no further charge, and nothing shuts them off remotely. Speaking new lines needs a working key.

How fast is a new line, really?

Our own measurement, not a third-party benchmark: 44 ms to the first audible sound for a Phonon voice, measured from one client in AWS us-east-1, one request at a time, with the voices streamed from our Mac mini over the public internet, November 21, 2026 (add your network time). The speed depends on your connection.

See all packs Pack documentation