audEERING

What is audEERING?

audEERING is a German company specializing in voice and audio artificial intelligence, with a focus on extracting non-verbal information from sound. Its core product, devAIce, is available in the form of SDKs, Web APIs, and XR plugins, and it is designed for use in software, hardware, call centers, automobiles, robots, wearable devices, as well as research projects.

It is not a website for audio editing, music generation, or general text-to-speech conversion. Its focus is on detecting sound activity, acoustic events, expressive dimensions, rhythm, speaker characteristics, audio quality, and the acoustic environment in which the audio is found.

The main products of audEERING

devAIce SDK

The SDK runs on local devices, making it suitable for scenarios that require low latency, offline processing, or the avoidance of continuous audio uploads. The official website lists Windows, Linux, macOS, Android, and iOS as supported operating systems, with support for both x86-64 and ARMv8 architectures.

devAIce Web API

Web APIs are used for batch audio analysis in the cloud; they are not graphical editors intended for ordinary users. Developers can obtain JSON results using the command line, direct HTTP requests, or clients written in Python, .NET, Java, JavaScript, and PHP.

devAIce XR

devAIce XR introduces voice expression and scene recognition capabilities to game engines such as Unity and Unreal. It can be used for virtual characters, immersive experiences, and player research; however, the specific modules and commercial licensing details need to be confirmed.

AI SoundLab

AI SoundLab is a development platform used for collecting audio data and conducting real-time analysis. In the context of health research, contributors can manage their own data; according to the official website, the relevant audio files and personal information are stored in Germany and can be deleted there.

openSMILE

openSMILE is a large-scale audio feature extraction tool maintained by audEERING, which can be used for analyzing speech, music, emotions, speakers, and rhythm. An official Python wrapper is also provided to enable researchers to easily extract common feature sets.

What can devAIce analyze?

BASE basic package

• Voice activity detection is used to distinguish between spoken segments and non-spoken segments.
• Acoustic event detection is used to identify sound events such as phone rings and crying.
• Speaker verification is used to compare the acoustic characteristics of a voice.
• Automatic speech recognition, which converts spoken words into text.
• Audio quality analysis and extraction of general acoustic features.

EXPRESSION expression package

• Voice expression and large-scale expression models.
• Multimodal expression analysis.
• Intonation, speech rate, and other rhythmic features.
• Continuous expression dimensions such as activation, pleasure, and dominance.

SPEAKER and SCENE

• Estimate speaker attributes such as age and perceived gender.
• Distinguish between three types of acoustic scenarios: indoor, outdoor, and traffic.
• Eight distinct scenarios are further identified under the three main categories.

In what scenarios is audEERING suitable?

• The call center analyzes the vocal expressions and the quality of conversations between customer service representatives and customers.
• Robots, cars, and smart devices adjust their interactions based on the sound environment.
• Games and XR applications enable virtual characters to respond to the player’s tone.
• The market research team analyzes the expressions found in interviews and product feedback.
• Research institutions extract reproducible audio features and train models.
• Medical research explores acoustic biomarkers, but they cannot be used directly for clinical diagnosis.

devAIce Web API Usage Guide

1. Determine whether the task requires the BASE, EXPRESSION, SPEAKER, or SCENE module.
2. Prepare audio material that is legally authorized, and obtain the informed consent of the speaker.
3. Purchase prepaid minutes using a small amount of non-sensitive samples, or contact sales to request a plan.
4. First, use the CLI recommended by the official sources to upload the file and check the JSON output.
5. Verify that the sampling rate, channels, noise level, and recording equipment meet the model’s requirements.
6. Use manually labeled samples to evaluate performance across different languages, accents, ages, and environments.
7. After confirming that the output meets the requirements, choose either a client library or direct HTTP integration.
8. Establish processes for retrying failed attempts, monitoring quotas, deleting data, and implementing access control.
9. Conduct assessments of privacy, labor relations, anti-discrimination policies, and regional regulations prior to going live.

Suggestions for integrating SDK

• Compare the processing speed, CPU usage, and memory usage on the actual target device.
• Voice activity detection is used to exclude non-voice segments that do not require analysis.
• Test under quiet, noisy, far-field conditions as well as with different microphones at the same time.
• Do not draw high-risk conclusions about individuals based on a single short audio clip.
• Determine whether analysis on the device side still requires an internet connection for license verification.

The advantages of audEERING

• It focuses on audio and non-verbal speech analysis, with a clear product orientation.
• It also provides device-side SDKs, cloud-based Web APIs, and XR plugins.
• It can run on desktop, mobile, and embedded architectures.
• The prepaid Web API is suitable for small-scale testing, as there is no need to sign a long-term subscription in advance.
• The authorities maintain openSMILE and a range of audio research tools.
• It supports clients in various programming languages as well as direct HTTP calls.

Usage restrictions and ethical risks

• Verbal expressions do not equate to a person’s true emotions, intentions, or personality.
• Age and perceived gender are model estimations that may be incorrect and reinforce biases.
• Accent, language, culture, disabilities, equipment, and the environment can all affect the outcome.
• Expression scores must not be used directly in high-risk decisions such as hiring, firing, insurance, or law enforcement.
• Sounds can constitute biometric or health-related data, and a legal basis must be established for their use.
• Health demonstrations are not medical products, nor do they provide medical advice.
• The prepaid API is a developer interface that does not include a full visual analysis workspace.

devAIce Web API pricing

The following are the prepaid options published on the official website as of August 26, 2026; they are suitable for testing and occasional analysis. Payments are processed through Stripe; fees, validity periods for minutes, additional modules, and any special benefits are specified on the purchase page.

The official website currently advertises that all packages are available for a limited time, but this promotion might end soon. There are no fixed prices listed for subscriptions, SDKs, XR plugins, commercial licenses, or customization consultations; it is necessary to contact sales for details.

Data and Privacy

The website’s privacy policy covers website visits, Cookies, contact forms, and marketing data, and it states that the GDPR is followed. Voice samples and customer content included in the products are also governed by the terms of service and specific contracts; therefore, it is not sufficient to rely solely on the website’s privacy page for information.

The terms of service require that customers possess the necessary rights and approvals for audio, sound samples, and other content, and that they assume the responsibilities outlined in GDPR. The terms also grant audEERING broad permission to use, copy, modify, and display customer data in order to fulfill the agreement, as well as the ability to utilize aggregated anonymized data.

Privacy check before deployment

• Determine what kind of consent is required for recording, uploading, authentication, and attribute inference respectively.
• Specify the retention period for the original audio, features, transcriptions, and inference results.
• Compare the data pathways for Web API upload processing and the device-side SDK.
• Restrict who can view individual-level expressions and speaker attributes.
• Establish processes for data access, deletion, objections, and manual review.

Is audEERING open source?

The devices, XR plugins, and business model of audEERING are not part of a fully open-source product. The official GitHub repository provides openSMILE, Python wrappers, audb, audmodel, and other research tools; however, it is necessary to check each individual project’s license.

The openSMILE source code is available for non-commercial uses such as personal use, research, and education, and it also permits commercial research under certain conditions. To use its code, derived versions, or extracted features in commercial products, a separate commercial license is usually required; it cannot be considered fully open-source just because it is available in public repositories.

Frequently Asked Questions

What does audEERING mainly do?

It focuses on voice and audio AI, using devAIce to analyze sound activity, expression, speaker characteristics, audio quality, acoustic events, and scenarios.

Can devAIce run offline?

The SDK can run on local devices; generally, there is no need to upload the analyzed audio to the cloud continuously. However, details regarding license checks and specific deployment requirements must be confirmed with the official party.

How much does the devAIce Web API cost?

The public prepaid packages are 35 euros for 600 minutes and 250 euros for 6,000 minutes; quotes are required for subscriptions, SDKs, XR, and enterprise licenses.

Can voice emotion recognition determine true feelings?

No. Model analysis focuses on the characteristics of vocal expression, and its results are influenced by language, culture, accent, and environment; it should not be used as a determination of an individual’s true emotions or intentions.

Can openSMILE be used commercially for free?

Generally, it is not allowed. The public version is intended for non-commercial, research, and educational purposes; commercial products or any derived works require a commercial license.

Is audEERING an open-source company?

No. The company makes several research tools available publicly, but its core commercial products, devAIce, as well as the models and services it offers, are still provided under commercial licenses.