Allinga TTS
Allinga TTS is a professional AI text-to-speech system from Fraunhofer IIS. It builds custom, multilingual, SSML-controlled voices for media, transport, alerting and industry, and runs either hosted in Germany or on your own server.
What is Allinga TTS?
Allinga TTS is an artificial-intelligence text-to-speech system built for professional environments. It is developed by Fraunhofer IIS, the Fraunhofer Institute for Integrated Circuits IIS in Erlangen, within its Audio and Media Technologies division, under the positioning AI-driven Speech Technologies for Professional Environments and the promise of AI solutions Made in Germany. Allinga is an umbrella brand that also covers Allinga AI Model Factory, a separate offer dedicated to custom large language models; that offer is not what this page reviews.
The system reaches its users two ways. The first is a subscription cloud service, operated and billed by LZE GmbH, an independent technology company based in Erlangen: it runs in a standard web browser with nothing to install, and it is hosted in Germany, which is what founds its claimed compliance with European data protection rules. The second is an installation on the company's own server, presented as delivering complete data sovereignty, GDPR compliance, availability with no constraint of time or place, and latency below 100 milliseconds, which the publisher describes as faster than real-time.
Four families of functions structure the product. Custom Voice clones an existing voice or creates a unique one, with its own accent and style, alongside a catalogue of premium voices. Multilingual Voices speak several languages and can switch from one to another in the middle of a sentence. Speech Control exposes rate, pitch and emphasis through standard SSML, driven by an in-house editor that works from the phoneme up to the full sentence, covering pronunciation, volume, pitch and speed. Emotions tint the delivery to match the context, enthusiastic for advertising, calming for an alert or a healthcare message. A customisable dictionary absorbs specialist terminology and in-house abbreviations, and the German dictionary receives continuous updates. The publisher claims clear, fluent delivery and correct pronunciation of international names and foreign words.
Four fields of application are documented: media production, service announcements, alerting solutions and industry. An interactive online demonstrator lets a visitor pick a scenario, among public announcements, media and alerts, and hear the result before talking to anyone. Historically, Allinga was developed jointly by the Fraunhofer IAIS and IIS institutes and also covered speech recognition; the product sheet of April 2024 still calls the speech module Allinga Voice.
What it does
- Generate professional-quality synthetic speech from written text
- Clone a voice, or build a bespoke one with its own accent and style
- Produce a single sentence in several languages, switching language mid-sentence
- Control rate, pitch and emphasis through SSML in a dedicated editor
- Colour the delivery with emotion: upbeat for advertising, calming for an alert or a health message
- Adapt the pronunciation dictionary to in-house jargon and abbreviations
- Install the engine on the company's own server to stay off the cloud, under 100 ms of latency
When to use Allinga TTS / When not to
A quick filter to help you decide if Allinga TTS is the right fit.
When to use Allinga TTS
- Radio stations, TV channels and content producers who have to put audio on air around the clock, including nights and weekends when no presenter is available
- Public transport operators and event organisers who need standardised, accessible service announcements
- Alerting and civil protection services generating warning messages in real time for public address systems and broadcast
- Manufacturers looking for hands-free human-machine interaction on the production line
- Organisations under GDPR that will not put their text through a foreign cloud: an on-site installation is possible, with latency below 100 ms
When not to use Allinga TTS
- Teams looking for an API to integrate: no API documentation is published, on any of the 22 pages checked or on the neighbouring domains
- Anyone expecting a mobile app or a browser extension: none exists
- Users who need speech recognition, that is speech-to-text: Allinga TTS only does synthesis
- Projects that require a catalogue of dozens of languages: four are named (German, English, French, Spanish), and the cloud offer advertises only three
- Buyers who need a documented contractual commitment on how their data is handled: no DPA and no subprocessor list is published for the product
How to use Allinga TTS
A typical end-to-end flow, from setup to results.
- Cloud route: start with the public demonstrator and listen to the scenarios to judge the voices for yourself
- Open the Allinga TTS page operated by LZE GmbH and compare the tiers against the volume of characters you expect to use each month
- Register with the LZE Subscription Manager and pick a subscription; Fraunhofer advertises 30 free days on the Basic plan
- Work straight from a standard web browser: there is no software to install
- Enter your text, choose a voice from the catalogue or commission a custom one, and generate the audio; Fraunhofer states that only a few minutes separate registration from the first speech output
- Fine-tune the delivery in the SSML editor: pronunciation, volume, pitch, speed, and language switching inside a sentence
- Extend the customisable dictionary with your own specialist terms and abbreviations
- On-site or bespoke route: use the Allinga contact form on the Fraunhofer IIS website to trigger a demonstration or an evaluation by the institute's experts
- The product sheet names Mandy Garcia, mandy.garcia@iis.fraunhofer.de, +49 9131 776-6178, as the product contact on the Fraunhofer side
- For technical questions about the running cloud service, write to contact@lze-innovation.de
Pros & Cons
Pros
- Real data sovereignty: the engine can be installed on the customer's own server, with no dependency on a cloud or on an internet connection
- The cloud service is hosted in Germany, with GDPR compliance claimed explicitly
- Backed by Fraunhofer IIS, the institute behind the audio coding work that led to the mp3
- Claimed command of German ahead of foreign providers, with continuous updates to the dictionary
- Latency announced below 100 ms in a local installation, compatible with real-time use
- Fine prosody control through an SSML editor designed for people who are not speech specialists
- Accessible entry point: EUR 15 a month, nothing to install, straight in a browser
Cons
- No published API: the engine cannot be industrialised without going through a bespoke project
- The cloud offer is sold and billed by a third party, LZE GmbH, whose terms and conditions never mention Allinga
- No document says what becomes of the submitted text: no ownership clause, no retention period, no subprocessor list, no position on model training
- No security certification is published, neither ISO 27001, nor SOC 2, nor TISAX, nor C5
- The language lists diverge from page to page: five variants on the Fraunhofer side (DE, EN, FR, DE-AT, ES), three on the LZE side (English, German including an Austrian accent, French), three again in the PDF product sheet
- The displayed prices carry a Winter Sale with no validity date, while the Fraunhofer page still promises an early bird discount
- The LZE pricing page links to a dead Fraunhofer URL, www.allinga.fraunhofer.de/en/contact.html, which returns 404
Pricing & Plans
There is no permanently free plan. The price list is published and billed by LZE GmbH, the third party that operates the cloud offer, and not by Fraunhofer; it is denominated in euros and includes German VAT at 19%, with private customers elsewhere in the European Union charged the rate of their own country, so the amount payable may differ. The entry point is the Basic subscription at EUR 15.00 per month, or EUR 150 per year, and the tiers are graded by volume of characters rather than by minutes of audio. The Premium and Advanced tiers are displayed with a Winter Sale reduction for which no start or end date is published, so the price actually charged should be checked at the moment of subscribing. A Business tier is offered without any published amount. Fraunhofer advertises a 30-day free trial on the Basic subscription, and no shipping or delivery charge applies.
- Basic — EUR 15 per month or EUR 150 per year — 100
- 000 characters per month
- 1.2 million per year — presented as the tier for an initial trial
- Premium — EUR 75 per month
- shown at EUR 52 in the Winter Sale
- or EUR 750 per year
- shown at EUR 520 — 500
- 000 characters per month
- 6 million per year — described as a starter pack for individuals or creators
- Advanced — EUR 150 per month
- shown at EUR 91 in the Winter Sale
- or EUR 1
- 500 per year
- shown at EUR 910 — 1 million characters per month
- 12 million per year — described as the tier for creators with bigger projects
- Business — no published amount — priority support and bespoke solutions
- arranged through Fraunhofer IIS
Data, GDPR & hosting
A consolidated view of how Allinga TTS handles your data.
GDPR overview
GDPR compliance is claimed three times — on the Fraunhofer product page, in the product PDF and on the LZE cloud page — and it rests on where processing happens: the cloud service is hosted in Germany, while an on-site installation is presented as giving complete data sovereignty. The publisher is established in Germany, so an Article 27 representative does not apply and none is designated. A data protection officer is reachable at datenschutz@zv.fraunhofer.de, and the controller is Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V., Hansastraße 27 c, 80686 München. The privacy policy itself is dated May 2018 and covers the institute's website: it details data subject rights under Articles 15 to 21 and highlights the Article 21 right to object, but no document describes the processing of an Allinga TTS user's data. No security certification is published.
Who owns the data?
No published document states who owns the text submitted to Allinga TTS or the audio it generates. The only privacy policy reachable from the product pages covers the website www.iis.fraunhofer.de, not the service: its controller is Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V., acting for Fraunhofer IIS, with a data protection officer at datenschutz@zv.fraunhofer.de. The product argument that an on-site installation delivers complete data sovereignty, GDPR compliance, and unrestricted availability in terms of time and location describes where processing happens; it is not an ownership clause. Anyone who needs one must obtain it contractually, from LZE GmbH for the cloud subscription or from Fraunhofer IIS for an on-site project.
Reuse rights
Nothing published says whether the end user may freely reuse what comes out of the engine, and nothing says what happens to the text that goes in. Two separate planes have to be kept apart here. For visitors to the institute's website, the policy is explicit: IP addresses are kept in the logs for four days, cookies for 90 days, event registration data for six months, and the named processors — Mailingwork for the newsletter, with servers in Germany, Wiredminds/LeadLab for audience analytics, and Friendly Captcha — handle that website traffic. An opt-out cookie is offered for audience analytics. None of those commitments is made on behalf of the product, and no equivalent opt-out is documented for model training. A Fraunhofer post of 1 August 2025 mentions gathering insights from user feedback to improve the technology, which is feedback rather than customer content. Reuse rights and permitted uses therefore have to be settled in the subscription contract with LZE GmbH, or in the agreement covering an on-site installation with Fraunhofer IIS, rather than read off the website.
Data retention & training
Hosting summary
Where the data sits depends on which route you take. The cloud service is stated to be hosted in Germany, an assertion repeated four times, in English and in German, on the LZE pricing pages, and that German hosting is exactly what founds the claimed compliance: "Allinga TTS is available as a cloud solution which complies with European data protection regulation, as it is hosted in Germany." With the alternative on-site deployment, the engine runs on the company's own server, so no data leaves its infrastructure at all and hosting becomes entirely the customer's responsibility. Beyond the country, nothing is specified: no city, no hosting operator and no data centre is named, and no transfer outside the European Union is mentioned in either direction. No security certification accompanies the hosting claim. A buyer who has to document a chain of custody will need to obtain those details directly, from LZE GmbH for the cloud service or from Fraunhofer IIS for an on-site installation.
Things to keep in mind
Risks and trade-offs to weigh before adopting Allinga TTS.
- The subscription is taken out with LZE GmbH (HRB 16647, Amtsgericht Fürth), not with Fraunhofer: check who your contract is actually with before you commit
- LZE's terms and conditions are those of its online shop and never mention Allinga; they address professionals, and ask the buyer to confirm its entrepreneurial status
- No document specifies what becomes of the text submitted to the cloud service, so nothing about retention, reuse or model training can be assumed: ask in writing
- The Winter Sale reduction and the early bird discount are displayed without any validity date: check the price actually charged at the moment of subscribing
- Quotas are expressed in characters, not in minutes of audio: size your real usage before choosing a tier
- The privacy policy reachable from the product is dated May 2018 and covers the institute's website, not the service you are paying for
- Voice cloning calls for its own rules: Custom Voice reproduces a real person's voice, so consent from the voice owner and an internal policy on where the clone may be used are worth settling before anything goes into production
Setup & Integrations
Technical difficulty
Two very different levels. The cloud route calls for no technical skill: nothing to install, everything runs in a standard web browser, you register with the LZE Subscription Manager, and Fraunhofer states that a few minutes separate registration from the first speech output. The interface is presented as usable without prior text-to-speech experience. Fine control goes through SSML, a standard syntax that takes some learning but comes with a dedicated editor. The on-site route is an integration project run with the Fraunhofer teams, with no public procedure or documented prerequisites. With no API, any automation means a bespoke project.
Deployment
Supported languages
Behind Allinga TTS
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Who develops Allinga TTS?
How do I get access to Allinga TTS?
How much does Allinga TTS cost?
Is there a free trial?
Which languages do the voices speak?
Where is the data hosted?
Is there an API?
Can I clone my own voice?
How do I control the intonation?
What latency can I expect?
Should you pick Allinga TTS?
Allinga TTS holds a clear position: a professional European text-to-speech system whose central argument is not the size of its voice catalogue but the place where the text is processed. That argument reaches two very different audiences through two channels. Creators, small teams and media producers subscribe to the browser-based cloud service, operated and billed by LZE GmbH from EUR 15 a month. Organisations handling sensitive material, alerting services, transport operators and industrial sites have the engine installed on their own server, which keeps everything inside their infrastructure and brings latency below 100 milliseconds. The structural strength lies in that combination: an applied research institute behind the technology, and a real option to keep the whole pipeline in-house instead of renting a foreign cloud.
The structural weakness is documentary. Product communication stops roughly where a buyer's questions start. There is no API, so any automation turns into a bespoke project. There is no data processing agreement, no subprocessor list and no published position on model training. Nothing says what happens to a text submitted to the cloud service, and the privacy policy reachable from the product covers the institute's website, in a wording dated May 2018, rather than the service itself. No security certification is published either.
One point deserves attention before signing anything: the value chain is split between three entities. Fraunhofer IIS develops the technology, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. is the legal entity, and LZE GmbH is the company a cloud customer actually contracts with. Knowing who answers for support, for billing and for data is part of the purchase. For a buyer whose priority is keeping speech synthesis on European ground, this is a serious candidate; for one who needs written guarantees, the conversation belongs before the subscription.
- Choosing a selection results in a full page refresh.
- Opens in a new window.