8 Voice Presets for Assistive Devices and Accessibility Solutions

By EchoEase editorial team · Updated 2026-11-19

This guide helps product teams building assistive devices and elderly care tools select the right voice preset from our eight options. We focus on clarity, predictable pricing, and voices that remain understandable in high-noise or low-bandwidth environments.

Readers will find practical criteria for matching voices to specific use cases like medication reminders or news updates. The guide covers technical constraints, testing strategies with real users, and cost implications. You will learn how to integrate these voices into your workflow using our straightforward API, ensuring a stable and transparent audio experience for end users.

8 Voice Presets for Assistive Devices and Accessibility Solutions

Frequently asked questions

Which voice is best for elderly users?

Q: Which voice is recommended for elderly users? A: Sage offers a calm, instructional tone, while Aria provides a warm, clear narrator style. Both are clear. You can test both presets with your specific hardware to determine which remains most understandable in your target noise environments.

Are all voices priced the same?

Yes. All eight voice presets use the same underlying engine models: tts-1 and tts-1-hd. Pricing is based solely on character count, not voice selection. You pay $10 per million characters regardless of whether you use Aria, Orion, or any other preset voice.

Can I mix voices in one session?

No. Each API request generates audio for a single voice. To mix voices, you must send separate requests for each segment and join the resulting audio files. The API does not support multi-voice generation within a single request payload.

What is the difference between Aria and Luna?

Aria is a warm, clear narrator voice, while Luna is soft and storytelling-oriented. Aria is better for instructional clarity, whereas Luna suits narrative content. Both are female presets with identical pricing and technical capabilities, differing only in vocal timbre and delivery style.

Is Orion too deep for some users?

Orion is described as deep and documentary-style. While deep voices can sometimes reduce clarity in noisy environments, this depends on your device’s audio output. You can adjust the speed between 0.25 and 4.0 to improve intelligibility if the base tone feels too low.

Do voices support Hindi equally well?

Yes. Any of the eight voices can speak any of the 23 supported languages, including Hindi. There is no quality difference between voices for Hindi. Set the language parameter to 'hi' to ensure correct pronunciation, as the API guesses language from script by default.

Are the voices based on real people?

A: Yes. The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0), derived from LibriVox public-domain audiobook readings. Each generated file carries an inaudible watermark marking it as synthetic speech. They are not custom-cloned from specific individuals.

Can I clone a voice for my device?

No. We do not offer custom voice cloning or allow you to upload your own voice. You must choose from the eight available presets: Aria, Nova, Luna, Sage, Felix, Milo, Atlas, or Orion. These are fixed presets provided by the API.

What happens if the voice changes?

Voices are static presets and do not change dynamically. If you need a different voice, you must send a new API request with the desired voice parameter. The voice selection is determined entirely by your request payload and remains consistent for that generation.

Is the audio format dependent on the voice?

No. The audio format is independent of the voice selected. You can choose from mp3, wav, opus, flac, or pcm for any voice. The format choice affects file size and compatibility, not the vocal characteristics or pricing of the generated speech.

Can I adjust the tone of a voice?

You can adjust the speed between 0.25 and 4.0, which affects delivery rate but not timbre. There are no emotion tags, SSML, or pitch controls. The voice’s inherent tone is fixed; you can only modify how fast or slow it is spoken.

Are the voices copyrighted?

A: The underlying recordings are from LibriTTS-R (CC BY 4.0) and LibriVox public-domain readings. Generated audio files are returned directly in the response and are not stored.

How do I choose between male and female voices?

Choose based on your user preference or brand tone. Female options include Aria, Nova, Luna, and Sage. Male options include Felix, Milo, Atlas, and Orion. All voices support all 23 languages equally. Test the voices with your target audience to see which is preferred.

Do voices have different latency?

No. All voices use the same engine models (tts-1 or tts-1-hd), so processing time is identical. Latency depends on request queue load and text length. Typical speed is roughly real-time, with short phrases taking a few seconds to generate.

How does the API handle long texts?

The API limits input to 4096 characters per request. For longer texts, you must split the text into parts, generate audio for each part, and join the resulting audio files. This ensures consistent quality and avoids exceeding the single-request character limit.

What languages are supported?

We support 23 languages: Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish. Any voice can speak any of these languages. Vietnamese, Indonesian, and Thai are not supported.

Criteria for Selecting a Voice

Sound wave illustration for: Criteria for Selecting a Voice

When building assistive devices for elderly users, voice selection directly impacts comprehension and retention. Clear articulation ensures that critical information, such as medication instructions or emergency alerts, is understood without cognitive strain. Tone warmth influences user comfort, while pacing determines how much information can be absorbed in a single listening session.

Poorly chosen voices can cause fatigue or confusion, especially in high-noise environments. By focusing on clarity, natural warmth, and adjustable speed, product teams can create audio interfaces that feel supportive rather than mechanical. This guide helps you evaluate each of our eight presets against these practical criteria.

Clarity and articulation

Articulation is the foundation of accessibility. Voices like Aria and Sage offer clear enunciation, making them ideal for delivering precise instructions or reading detailed text. For users with hearing impairments, consistent consonant definition reduces the effort required to parse speech.

Voices have different timbres, but the API does not specify which ones cut through noise better than others. Consider your device's typical environment—whether it is a quiet bedroom or a busy living room—when prioritizing articulation.

  • Choose voices with distinct consonant sounds for medication reminders.
  • Avoid overly deep voices if the device operates in noisy areas.
  • Test audio output at maximum volume to ensure no distortion.

Tone warmth

Warmth affects emotional connection and long-term usability. A voice that sounds too robotic can feel impersonal, while one that is too casual might seem unprofessional for health updates. Aria provides a warm, clear narrator tone that balances friendliness with authority.

Milo offers a friendly, conversational style suitable for daily companionship or news updates. Sage delivers a calm, instructional tone ideal for step-by-step guidance. Selecting the right warmth ensures users feel respected and engaged rather than merely informed.

Consistency in tone builds trust. Once a voice is chosen for your device’s interface, stick with it across all features to avoid confusing users who have grown accustomed to a specific auditory identity.

  • Aria: Warm and clear, good for general assistance.
  • Milo: Friendly and conversational, ideal for daily updates.
  • Sage: Calm and instructional, best for guided tasks.

Pacing and speed

Natural pacing varies by voice and use case. Sage and Aria tend to have a more measured rhythm, which can aid comprehension for slower listeners. Nova and Felix speak with more energy, which might feel too fast for some elderly users unless adjusted.

The API allows speed adjustment from 0.25x to 4.0x. For assistive devices, reducing speed to 0.8x or 0.9x often improves clarity without making the speech sound drawn out. Experiment with these settings to find the sweet spot for your target demographic.

Remember that longer texts are split into parts of up to 4096 characters and joined. Ensure your UI handles these transitions smoothly, maintaining the chosen pace and tone across segments to prevent jarring audio breaks.

  • Use speed 0.8–0.9x for elderly users with slower processing speeds.
  • A/B test speeds with real users to find optimal clarity.
  • Ensure seamless transitions between 4096-character text segments.

Female Voices Overview

Sound wave illustration for: Female Voices Overview

The four female presets offer distinct tonal profiles suited for assistive interfaces. Each voice is clear, but intelligibility in noise depends on the environment and volume.

Product teams can select a preset based on the desired user experience, whether it requires a comforting presence for long-form content or a clear directive for quick alerts.

Aria: Warm narrator

Aria provides a warm, clear narrative tone ideal for reading books or detailed instructions. Her delivery is steady and easy to follow for elderly users who may have mild hearing difficulties.

  • Best for: Audiobooks, step-by-step guides, and general assistance.

Nova: Bright upbeat

Nova delivers a bright, energetic performance that keeps engagement high. She is well-suited for news updates or short reminders where a lively tone prevents the interface from feeling robotic.

  • Best for: News briefings, weather alerts, and motivational prompts.

Luna: Soft storytelling

Luna offers a soft, gentle voice that reduces cognitive load during storytelling or relaxation sessions. Her pacing is naturally soothing, making her a strong choice for evening routines or sleep aids.

  • Best for: Bedtime stories, calming sounds, and gentle reminders.

Sage: Calm instructional

Sage provides a calm, instructional tone that prioritizes clarity over emotion. She is effective for medication reminders or health tips where precise information delivery is critical and background distraction is high.

  • Best for: Medication alerts, health tips, and instructional commands.

Male Voices Overview

Sound wave illustration for: Male Voices Overview

The male voice presets offer distinct tonal profiles suitable for various assistive device contexts. Each voice has a distinct tone, but none are specifically engineered for high-noise environments.

Felix: Energetic promo

Felix delivers an energetic tone, ideal for promotional updates or quick alerts. His clear articulation ensures that time-sensitive information is communicated effectively without fatigue.

  • High energy delivery
  • Clear articulation for alerts

Milo: Friendly conversational

Milo provides a friendly, conversational tone that feels natural during extended interactions. This preset is particularly effective for companionship features or daily check-ins, reducing the mechanical feel of synthetic speech.

  • Natural conversational flow
  • Suitable for long-form interaction

Atlas: Neutral news

Atlas offers a neutral, news-style delivery. This voice minimizes emotional bias, making it ideal for factual updates, weather reports, or medication reminders where clarity and neutrality are prioritized over personality.

  • Neutral and unbiased tone
  • High clarity for facts

Orion: Deep documentary

Orion features a deep, documentary-style voice. The lower pitch can be easier for users with high-frequency hearing loss to distinguish, providing a stable and grounding auditory experience for audiobooks or guided meditations.

  • Deep, lower pitch
  • Stable delivery for content

Matching Voice to Use Case

Sound wave illustration for: Matching Voice to Use Case

Matching the right voice to your assistive device’s primary function improves user compliance and reduces cognitive load. By aligning vocal characteristics with specific use cases, product teams can create more intuitive and effective auditory experiences for elderly users and those with hearing impairments.

Medication reminders

Use Sage for calm, instructional reminders or Milo for a friendly, conversational tone. These presets ensure clear enunciation without sounding robotic, helping users remember tasks without feeling rushed.

  • Sage: Calm, instructional tone ideal for step-by-step guidance.
  • Milo: Friendly, conversational style for daily check-ins.

Book reading

Aria provides a warm, clear narrator voice, while Luna offers a soft, storytelling quality. Both support longer texts up to 4096 characters, making them suitable for audiobooks and extended reading sessions where clarity is paramount.

  • Aria: Warm, clear narrator for standard reading.
  • Luna: Soft, storytelling voice for immersive experiences.

News updates

Atlas delivers a neutral, news-style voice perfect for factual updates. Its clear articulation ensures important information is understood quickly, even in noisy environments common in communal living spaces.

  • Atlas: Neutral, news-style voice for factual updates.

Voice Characteristics for Elderly Users

Sound wave illustration for: Voice Characteristics for Elderly Users

Elderly users often experience presbycusis, a gradual loss of high-frequency hearing that makes consonants difficult to distinguish. Selecting voices with appropriate pitch and pacing ensures that critical information remains audible without requiring excessive volume adjustments.

Lower pitch preference

High-frequency sounds attenuate faster in aging ears. Male voices like Atlas and Orion have a lower pitch, but the API does not claim they cut through noise more effectively than female voices.

Slower natural pace

Standard speech rates can overwhelm cognitive processing speed. Reducing the speed parameter to 0.25–4.0 allows users more time to parse words. Sage (calm, instructional) and Luna (soft, storytelling) naturally lend themselves to slower, deliberate delivery.

Clear enunciation

Articulation clarity reduces listening effort. Aria (warm, clear narrator) and Milo (friendly, conversational) offer distinct consonant transitions. Test these presets with medication reminders to ensure phonemes remain distinguishable at lower volumes.

Technical Constraints

Sound wave illustration for: Technical Constraints

Understanding the technical boundaries of the EchoEase API ensures your assistive device delivers audio without interruption. Knowing these limits prevents unexpected failures during critical interactions, allowing product teams to design robust fallback mechanisms for text processing and playback.

Character limits

Each API request accepts up to 4096 characters, including spaces and punctuation. For longer texts, your software must split the input into chunks and concatenate the resulting audio files. Short phrases typically generate within a few seconds, while longer texts process at roughly real-time speed.

  • Maximum input: 4096 characters per request
  • Splits required for longer narratives
  • Includes spaces and punctuation in count

Format compatibility

The API supports MP3, WAV, OPUS, FLAC, and PCM outputs. Choose MP3 or OPUS for bandwidth-efficient delivery in low-bandwidth environments, or WAV/FLAC for high-fidelity playback in quiet settings. PCM is ideal for raw audio processing.

  • MP3, WAV, OPUS, FLAC, PCM supported
  • Use OPUS for low bandwidth
  • WAV/FLAC for high fidelity

Speed range impact

Adjustable from 0.25x to 4.0x speed allows optimization for elderly users who may prefer slower pacing. Slower speeds enhance clarity but increase processing time. Faster speeds reduce latency but may reduce comprehension in high-noise environments.

  • Range: 0.25x to 4.0x
  • Slower speeds improve clarity
  • Adjust based on user preference

Testing Voices with Real Users

Sound wave illustration for: Testing Voices with Real Users

Selecting the right voice for assistive devices requires empirical validation with the end-users, particularly elderly individuals who may experience hearing or cognitive declines. Generic voice preferences often diverge from actual usability in high-noise or low-bandwidth environments, making direct observation essential.

Feedback collection

Gather structured feedback after users listen to generated audio. Focus on clarity, warmth, and ease of understanding rather than subjective preference. Ask specific questions about whether instructions were followed correctly and if the tone felt appropriate for the context.

  • Ask users to rate clarity on a 1-5 scale after each prompt.

Usability testing

Conduct real-world scenarios where users interact with the device. Observe if they can distinguish the voice from background noise. Test various speeds (0.25–4.0) to find the natural pace for slower processors or users with slower processing speeds. Record completion rates and error counts.

  • Test in environments mimicking home noise levels.

Iterative selection

['Refine your choice based on observed performance, not just feedback. If a voice like Aria is preferred but causes confusion in complex tasks, switch to Sage for clearer instructional tones. Validate each change with a new round of user testing before finalizing the preset for production.']

bullets

Testing Voices with Real Users

Cost Implications of Voice Choice

Sound wave illustration for: Cost Implications of Voice Choice

Voice selection does not affect your billing rate. Every preset, from Aria’s warm narration to Orion’s deep documentary tone, costs exactly $10 per 1 million characters. This uniformity simplifies cost forecasting for assistive device teams who may switch voices based on user preference without worrying about variable pricing tiers.

Uniform pricing

Regardless of whether you use tts-1 or tts-1-hd, the character count remains the sole metric for billing. A 4,000-character medication reminder costs the same as a 4,000-character news update, ensuring predictable operational expenses.

No voice-based costs

There are no premium fees for specific vocal qualities or languages. You can test multiple voices for the same elderly user cohort without incurring additional per-voice charges, allowing for flexible experimentation.

Budget neutrality

Since pricing is strictly character-based, switching from a fast-paced voice to a slower, more deliberate one does not increase costs unless the slower pace extends the total audio duration significantly. Generally, voice choice is budget-neutral.

Licensing and Data Usage

Sound wave illustration for: Licensing and Data Usage

EchoEase voices originate from the LibriTTS-R dataset, derived from public-domain audiobook readings on LibriVox. This foundation ensures that our synthetic voices are built upon licensed, high-quality human recordings rather than proprietary studio sessions.

Each voice preset is based on these licensed recordings, providing a consistent and legally clear audio source for your assistive devices. The data pipeline is straightforward: input text is synthesized and returned, with no retention for training purposes.

LibriTTS-R source

Our voices are generated from the LibriTTS-R dataset, which contains hour-long audiobooks read by volunteers. This dataset provides a diverse range of natural speech patterns, ensuring clarity and natural rhythm in the final audio output for elderly users.

CC BY 4.0 license

The underlying recordings are available under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. This allows us to use the recordings as a base for our TTS engine without additional licensing fees per character or voice, keeping your costs predictable.

  • Open commercial use allowed
  • Attribution required for the original dataset
  • No extra royalties per audio file generated

Public domain readings

The source material comes from LibriVox, a community of volunteers reading books that have entered the public domain. This ensures that the base audio content is free from complex copyright restrictions, reducing legal risk for your product team deploying assistive devices.

Watermarking and Synthetic Detection

Sound wave illustration for: Watermarking and Synthetic Detection

Every audio file generated by EchoEase contains an inaudible watermark marking it as synthetic speech. This ensures clear identification without affecting the listening experience for users of assistive devices or elderly care tools.

The watermark is embedded directly into the audio data, allowing for reliable detection by downstream systems or verification tools without requiring additional metadata files.

Inaudible watermark

The watermark is embedded directly into the audio data, ensuring reliable detection without affecting the listening experience. It remains consistent across all supported formats, including MP3, WAV, and OPUS.

  • Embedded in all generated audio files
  • Format-agnostic detection
  • No impact on audio quality

Synthetic identification

Systems can verify synthetic origin using standard detection methods. This helps maintain transparency in assistive devices where users need to distinguish between human-recorded and AI-generated content.

Detection tools can process the audio in real-time or post-generation, depending on your device's computational capacity.

  • Real-time or post-generation verification
  • Compatible with standard detection algorithms
  • No extra processing overhead required

Transparency in use

While not legally required, we recommend informing users that content is AI-generated. This builds trust and meets accessibility standards for synthetic media disclosure.

The watermark provides a technical guarantee of synthetic origin, complementing your transparency efforts.

  • Supports accessibility disclosure standards
  • Builds user trust
  • Technical verification alongside user communication

Final Recommendations for Product Teams

Sound wave illustration for: Final Recommendations for Product Teams

For assistive devices, begin with Aria or Milo. Aria provides a clear, warm narrator tone ideal for general instructions, while Milo offers a friendly, conversational style that feels less robotic for elderly users. Both presets are clear, but clarity in noisy environments depends on the device's volume and environment.

Use A/B testing with real users to validate these choices. Present both voices in parallel during usability trials to measure comprehension and comfort levels. This data-driven approach helps teams select the most effective voice for their specific demographic before committing to a single preset.

Start with Aria or Milo

Begin integration with Aria or Milo as your default presets. Aria’s warm, clear narration is excellent for general assistive commands, while Milo’s conversational tone reduces the "robotic" feel often reported by elderly users. Both voices demonstrate strong articulation, which is critical for users with hearing impairments or cognitive load.

  • Aria: Best for clear, structured instructions.
  • Milo: Ideal for conversational, friendly interactions.

A/B testing strategies

Implement A/B testing by rotating voices in your live environment. Track user completion rates and feedback for each voice variant. Since voices are consistent presets, you can directly compare user preference and comprehension without introducing variable audio quality factors.

Long-term voice stability

Once a voice is selected, maintain it for consistency. Unlike custom cloning, these eight presets are stable and do not change over time. This ensures that users do not experience voice shifts as your product updates, maintaining trust and familiarity in assistive interactions.

API Integration for Assistive Tech

Sound wave illustration for: API Integration for Assistive Tech

Integrate EchoEase into your assistive device firmware or companion app using a standard JSON endpoint. Send a POST request to the speech API with your text, selected voice, and format preferences. The system returns the generated audio file directly, enabling seamless playback in your user interface without complex streaming logic.

Simple JSON endpoint

Use the standard POST endpoint to send text up to 4096 characters. Specify the model, voice preset, and desired audio format (mp3, wav, etc.) in the JSON body. For longer content, split the text into chunks, generate each part, and concatenate the resulting audio files in your application logic.

  • Endpoint: POST https://api.bethkd.info/v1/audio/speech
  • Supports mp3, wav, opus, flac, and pcm formats
  • Text input limited to 4096 characters per request

OpenAI SDK compatibility

The API mirrors the OpenAI speech endpoint structure. You can use the official OpenAI SDKs by simply changing the base_url to https://api.bethkd.info/v1 and providing your EchoEase API key. This reduces integration time for teams already familiar with common TTS patterns.

  • Compatible with official OpenAI SDKs
  • Change base_url to https://api.bethkd.info/v1
  • Use Authorization: Bearer tts-... for authentication

Rate limits and scaling

Each API key allows 60 requests per minute once funded, or 10 requests per minute during the trial phase. Requests are processed sequentially, so high-volume deployments may experience queueing delays. Plan your audio generation timing accordingly, especially for real-time feedback features in assistive devices.

  • 60 requests/minute per key (standard)
  • 10 requests/minute (trial phase)
  • Sequential processing may cause queueing under load

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. This helps maintain transparency in assistive devices where users need to distinguish between human-recorded and AI-generated content.

Avatar of the voice aria
ariafemale
Avatar of the voice nova
novafemale
Avatar of the voice luna
lunafemale
Avatar of the voice sage
sagefemale
Avatar of the voice felix
felixmale
Avatar of the voice milo
milomale
Avatar of the voice atlas
atlasmale
Avatar of the voice orion
orionmale

23 languages

Any voice speaks any of these languages.

Sources and further reading

  1. Wikipedia
  2. openai-node on GitHub
  3. openai-python on GitHub
  4. W3C WAI: making audio and video accessible
  5. Node.js documentation
  6. FFmpeg documentation
  7. Python documentation
  8. Tether (USDT)
  9. curl documentation
  10. LibriVox public-domain audiobooks
  11. OpenAI API reference: create speech
  12. IETF RFC 7845: Ogg Opus

More on this site