End User License Agreement

This End User License Agreement (EULA) specifically applies to the WritingAnt software (Product Brand Name: WritingAnt, Company Name: Ögren Digital) that users download and install on their computers (with functionality related to speech-to-text, text-enhancement, text-generation, voice commands for computer control). By downloading, installing, accessing, or using the Software, you agree to be bound by the terms and conditions set forth below.

1.1 The Software and related Services are provided solely to individuals (consumers) who are residents of, and business entities (including sole proprietors, companies, and organizations) that are incorporated or legally established in and primarily operating within, Switzerland, Germany, France, Italy, Ireland, Sweden, Portugal, or Spain (together, the "Eligible Territory").

1.2 By downloading, installing, accessing or using the Software, you represent and warrant that you meet the above residency or incorporation requirement. If you cease to qualify, you must immediately discontinue all use and uninstall/delete any local copies. The Company may geo-block or suspend access at its discretion to enforce this clause.

2.1 Your right to process audio or text through the Service is subject to the monthly request quotas applicable to your active subscription plan as described on our Subscription Plan page.

2.2 Hard Usage Limits: The following monthly limits apply based on your subscription tier:

2.3 Hard Limit Reached: Once the Quota is exhausted, further requests will be rejected until the next billing cycle unless you upgrade. Excess usage attempts via fraudulent behavior may incur overage fees as listed on the Pricing page.

2.4 We reserve the right to throttle or suspend accounts exhibiting abnormal traffic patterns to protect service stability.

2.5 Maximum Text Length per Editing Request: Each request to edit selected text with the writing assistant may contain up to approximately three pages of text (10,000 characters). This limit applies to all subscription tiers and is measured on the readable text of your selection only — formatting markup is not counted. Selections exceeding the limit are declined by the application before any content is sent for processing, and no usage credit is consumed.

Payment Options for Subscription Plans
We offer several payment methods through our third-party provider, Stripe, as outlined in our Terms and Conditions at www.oegren-digital.ch/terms. Additionally, you may choose to pay by invoice via bank transfer. All subscription plans are billed on a recurring basis and renew automatically unless cancelled in accordance with the terms stated below.

Invoice Subscription Terms and Cancellation
If you choose to pay by invoice for a recurring subscription, the same terms regarding automatic renewal and cancellation apply as set forth in our Terms and Conditions. You may cancel your subscription at any time by providing written notice (e.g., via email) no later than the last business day before the start of the next billing period. Failure to pay an invoice by the due date may result in suspension or termination of your access to the service. No additional fees apply for cancellation.

Failed Payments and Account Suspension
If we are unable to successfully process your payment using the selected method (e.g., due to insufficient funds, expired card, or rejection), we may attempt to retry the payment. If full payment is not received within seven (7) calendar days of the original due date, we reserve the right to suspend or terminate your access to the service until the outstanding balance is paid. We will notify you before taking such action.

Taxes, VAT, or equivalent sales tax will be calculated at checkout based on your billing country. For customers choosing the optional bank transfer (invoice) payment method, taxes will also be included during the invoice preparation process.

Supported Languages for Speech/Writing
The dictation/writing features support multiple languages with varying levels of accuracy. Great speech/writing accuracy is available for: English (American and British variants), Spanish, German, French, Italian, Portuguese, and Russian. Medium accuracy is available for: Swedish, Danish, Norwegian, Dutch, and Ukrainian. Basic accuracy is available for: Polish, Czech, Slovak, Hungarian, Romanian, Croatian, Bulgarian, and Finnish.

Internet Connectivity Requirement
The Software requires an active internet connection at all times to operate. This connection is used to verify licence entitlement, subscription status, account integrity and usage quotas. “Offline speech‑to‑text” means audio is processed locally and no audio data is sent to the cloud; however, the application itself must still be online for the features to work.

Application Aware Preset Switching
The Software includes an optional feature ("Application Aware") that automatically detects the active application and, where you configure browser rules, the address of the active browser tab, in order to apply language and writing-style presets you define. This detection happens entirely on your computer; no application names or web addresses are transmitted to us. You must explicitly enable the feature in Settings and configure which applications and websites map to which presets. While it is disabled, no such detection is performed.

The performance of how fast transcription happens depends on the user's level of CPU and how many CPUs exist as well. On Mac computers, it also depends on the AI-tailored chip. On Windows computers, GPU acceleration may also be used when a compatible GPU is available.

We currently use GPT-5.6 with a fallback to GPT-4.1 if needed for availability. We may switch between them depending on availability and suitability for the application. These models may change in the future to whatever best suits the application as we continuously optimize for quality and performance.

6.1 Users may only access the Services through the installed Software provided by Ögren Digital. Direct API calls, reverse engineering of API endpoints, or any attempts to bypass the Software interface are strictly prohibited.

You may submit only content that complies with all of the following:

We may block or delete non-compliant content and may suspend or terminate the account of any user who violates this clause.

To verify compliance with Sections 1–7 and Section 11, the Company may (i) monitor aggregate usage metrics, and (ii) request written confirmation of your residency or legal incorporation. Failure to cooperate constitutes material breach.

Any breach of Sections 1–7 or Section 11 is deemed material. On notice (or automatic detection) of such breach we may immediately suspend or terminate your licence and/or subscription without refund.

Sections 6–14 survive termination.

AI Processing
We use AI for both local offline transcription and for text improvements, which are sent online to different third-party providers and models. We may engage trusted third-party AI providers to perform transcription, text generation, and text editing features. We may change between different AI models as necessary to maintain service quality and availability.

AI Limitations
Artificial intelligence, by its nature, may produce inaccurate, incomplete, or misleading outputs, including hallucinations or factual errors.

User Review Required
You are responsible for all content submitted, modified, or distributed via our Service. You must review and verify every AI-generated or AI-enhanced output before use. The Company does not guarantee the accuracy, reliability, or suitability of AI outputs.

Liability Exclusion (B2B & B2C)
Except for mandatory statutory rights:

Consumer Rights Preserved
If you are an EU consumer, this clause does not limit or exclude your mandated legal rights under the Digital Content Directive or related EU consumer protection laws (e.g. right to repair, replacement, or compensation for defects).

Disclosure & Transparency
In accordance with EU AI Regulation Article 50, we inform you that AI is used to generate or modify text. AI may introduce errors or artifacts, and you are notified accordingly. You must validate all AI output before use.

AI Models for Transcription
The AI models used for transcription on your local computer depend on your hardware capacity and software environment. We recommend testing your setup with the free tier to ensure you achieve the quality and speed needed for the languages you use most before upgrading to a paid tier. This way, you can confirm the system works well for your specific requirements.

AI Model Storage Location
When the AI transcription models are downloaded separately rather than bundled with the installer, they are stored locally on your computer in a dedicated folder. On macOS this is ~/Library/Application Support/SpeechToWellWrittenTextAssistant/ai_models, and on Windows it is %LOCALAPPDATA%\OegrenDigital\SpeechToWellWrittenTextAssistant\ai_models. You may delete this folder at any time to free up disk space; the Software will offer to download the models again when they are needed.

Powerful Tool with AI-Based Risks
The voice detection mechanism that starts and stops speech-to-text functions and controls computer operations or virtually presses keys is a powerful and useful tool. However, it relies on artificial intelligence technology, which can make mistakes.

User Responsibility for Appropriate Use
You are responsible for deciding when it is appropriate to use voice activation features and must accept that mistakes can happen. These mistakes should not have consequences greater than your ability to manage, detect, and undo them.

Risk Assessment Factors
The level of risk depends on several factors, including the type of documents you are working with, the importance of the content, and your ability to quickly detect and correct errors. You must evaluate these factors before using voice activation.

Environmental Considerations
Background noise, other voices, or music significantly increase the chance of errors from voice command activation. In noisy environments, you must carefully consider whether it is suitable to use voice activation features.

Alternative Safe Operation Mode
The voice command function can be easily disabled in the user interface, providing a safer way to work when needed. When disabled, you can start and stop speech-to-text functions by clicking buttons instead of using voice commands.

UI-Based Review Process
When operating without voice commands, text results appear in the user interface rather than directly at the cursor position in other applications. This allows you to review the text before using it elsewhere, which is especially helpful in noisy environments or when working with sensitive content.

User Accountability
You are responsible for making informed decisions about when and how to use these tools. With the power of voice activation comes the responsibility to use it appropriately and safely.

One Free Account Per Individual
Each individual person is permitted to maintain only one user account with a free subscription plan. This limitation is designed to prevent abuse of the free usage tier and ensure fair access to evaluation resources.

Free Tier Purpose
The free tier is intended for limited evaluation of the Software before purchasing a paid subscription or for individuals who want to have a low usage volume.

Multiple Paid Accounts Permitted
While each individual may have only one free subscription plan, you are permitted to maintain multiple paid subscription accounts. For example, you may have a private account for personal use and a separate account provided by your employer for business use.

Enforcement
We may monitor account creation patterns and usage to enforce this policy. Violation of this restriction may result in suspension or termination of all associated accounts without refund.

14.1 We provide user support exclusively by email at: support@oegren-digital.ch.

14.2 We do not offer support via telephone, live chat, or in-person assistance.

14.3 We aim to respond to inquiries within a reasonable timeframe during business hours (CET), but do not guarantee a specific response time or resolution outcome.

14.4 Support is available in English, German, Italian, and Swedish.

The local speech to text process on your computer uses your CPU and adapts with the amount of cores to engage based on your computer capabilities. On a modern computer, the transcription speed on CPU will be rather fast, but on older computers it can be slower. On Mac M1+ you can choose whether or not you want to use only the CPU or also engage the ANE Hardware Acceleration which is enabled by default which gives the fastest results for transcription. On Windows, the Software can also utilize GPU acceleration when a compatible GPU is available, which can significantly improve transcription speed.

Company News and Product Information
The Software may display company news, product announcements, service updates, feature releases, and other relevant information about our products and services within the user interface. This content is designed to keep you informed about improvements, new features, and important updates to enhance your experience.

Partner Offers and Promotions
We may occasionally display offers, promotions, or information about products and services from our trusted partners that we believe may be of interest to you. These partner offerings are carefully selected and relevant to our user base and the functionality of our Software.

Display Locations
Such news, announcements, and offers may appear in various locations within the Software interface, including but not limited to: startup screens, notification areas, settings panels, help sections, or dedicated news/announcements sections. Content will be clearly identified and will not interfere with the core functionality of the Software.

User Control and Preferences
You may have options to control the display of certain types of announcements or partner offers through the Software settings. However, critical security updates, service notices, and important product information may be displayed regardless of your preference settings to ensure proper operation and security.

Privacy and Data Use
The display of news and offers does not require transmission of additional personal data beyond what is already collected as part of normal Software operation. Partner offers do not result in sharing of your personal information with third parties unless you actively choose to engage with such offers.

No Additional Charges
Viewing company news, product information, or partner offers within the Software interface does not incur additional charges beyond your existing subscription fees. Any partner offers that involve separate purchases or subscriptions will be clearly disclosed with their own terms and pricing.

Visual Recording Indicator
During audio recording and processing operations, the Software will display a transparent yellow overlay covering the entire screen. This overlay serves as a clear visual indicator that recording or processing is in progress and helps prevent accidental interruption of these operations.

Input Blocking During Operations
While the yellow overlay is active, mouse clicks and other input interactions outside of the Software's recording interface will be blocked. This prevents accidental text insertion or application switching that could disrupt the recording or processing workflow and ensures accurate results.

User Safety and Experience
The overlay system is designed to enhance user experience by providing clear feedback about system status and preventing common user errors during critical operations. Users can still interact with the Software's recording controls and interface elements during this time.

Metrics Collection and Storage
We calculate and store writing-session metrics, including character counts, estimated keystrokes, and timing data. We utilize these metrics to estimate the time saved when using voice dictation compared to keyboard typing.

Time Savings Estimation
We also estimate the monetary value of this time saved. You may adjust the parameters applied to these calculations, including your assumed hourly value of time, your keyboard typing speed, and your speaking speed.

Two Ways to Use the Writing Assistant
You can use the Software in two ways. First, you can edit text that already exists: you select text in an application, and the text you selected is sent to the writing assistant and then replaced by the returned result. Second, you can use voice writing (dictation): you speak, your speech is converted into text, and you may optionally choose to apply automatic writing-style corrections that are applied after dictation, before the result is delivered and inserted. Both ways use the writing assistant and consume the same usage credit as described on the Subscription Plan page.

Selected Text Capture via Clipboard
When you operate the device and use the correct voice command, keyboard shortcut, or another trigger you have configured (such as an assigned mouse button or, on Mac, the Globe (fn) key), the Software will take the text you have selected in an application and automatically copy it to the operating system clipboard — or cut it (copy it and remove the selection) when you use the selected-text editing shortcut. It will then read the text from the clipboard and send it to the cloud for further processing. If a cut selection cannot be processed, the Software attempts to paste your original text back automatically.

Similarity to Recording-Based Processing
This feature works similarly to making a recording to capture text as a basis for further editing. The difference is that instead of recording speech, you select existing text. After processing, the system will automatically paste the result through your operating system.

Automatic Replacement of Selected Text
If the text is still selected in a third-party application at the time the processed result is returned, it will be replaced with the processed result. This replacement is performed as a cut followed by a paste: the selected text is first cut (removed) and the processed result is then pasted in its place through your operating system. The Software does its best to preserve rich formatting (such as headings, bold text, and lists) during this replacement, but depending on the target application, some formatting may be slightly adjusted, added, or removed. You are responsible for reviewing and verifying the processed output before continuing to work with it, as described in Section 11.

Clipboard Restoration and Its Limits
Because the Software inserts text by placing it on the operating system clipboard and then pasting it, an insertion would normally replace whatever you had copied to the clipboard before. To avoid this, the Software by default takes a snapshot of your current clipboard content (text, formatting, and images, where technically possible) immediately before an automatic insertion and puts it back shortly after the insertion has completed. The intention is that your own copied content survives dictation and text editing. If you prefer that the most recently inserted text — the result produced by the Software — stays on the clipboard after each insertion, for example so that you can paste it again elsewhere, you can turn this restoration off in the application settings. Restoration is performed on a best-effort basis: it depends on the timing behavior of your operating system and of the target application, and despite built-in safeguards it can occasionally fail. In particular, your previous clipboard content may not be restored (for example, copied files and certain application-specific clipboard content cannot technically be restored and remain replaced), or a paste may not deliver the content you expect (for example, when an unusually slow application reads the clipboard only after the restoration has already taken place). You should therefore not rely on the clipboard as the only place where important content is kept, and you remain responsible for reviewing inserted results, as described in Section 11.

Purpose of Local Folders
The Software and its helper processes (the launcher, the overlay user interface, and the local speech-recognition subprocesses) create a number of folders and entries on your computer to store settings, encrypted user data, AI models, caches, and diagnostic information. Data described as encrypted is encrypted at rest; the encryption key is stored in the operating system's secure key store (macOS Keychain or Windows Credential Manager, entry name "WritingAnt") and never leaves your device.

Main Data Folder
On macOS: ~/Library/Application Support/SpeechToWellWrittenTextAssistant. On Windows: %LOCALAPPDATA%\OegrenDigital\SpeechToWellWrittenTextAssistant. This folder contains your encrypted application settings, login session, dictation history, personal snippets, custom voice commands, custom writing styles, and feature preferences; optionally saved encrypted audio recordings (audio_recordings); downloaded AI transcription models (ai_models, not encrypted — see Section 11.9); crash and diagnostic reports (CrashLogs); application log files (logs); and non-sensitive startup and login status files. A development subfolder with the same structure is only used by development builds.

Temporary Files
On both operating systems the Software additionally uses the system's temporary folder for an instance lock file (writing_ant_main.lock) and for short-lived audio buffers during transcription; these temporary files are removed automatically.

Removal of Local Data
Deleting these folders or uninstalling the Software removes all locally stored data. If the ai_models folder is deleted, the Software will offer to download the AI models again when they are needed, as described in Section 11.9.