Frequently asked questions
Which voice is the most neutral?
Atlas is described as neutral and suitable for news, making it a reliable choice for legal documents. It provides a clear, straightforward delivery without strong emotional coloring. This neutrality ensures that complex case files and contracts are read back accurately, allowing legal teams to focus on the content rather than the narrator's personality.
Can I use Luna for legal briefs?
Yes, Luna is a soft, storytelling voice that can read legal briefs. While designed for storytelling, it supports all 23 languages and can handle the text input. Legal teams may prefer Atlas for strict neutrality or Sage for instructional clarity, but Luna remains a valid option for reading briefs with a softer tone.
Is Felix too energetic for contracts?
Felix is described as energetic and promo-oriented. For contracts, this high energy might feel less formal than Atlas or Sage. However, the choice depends on your team's preference. You can adjust the speed between 0.25 and 4.0 to moderate the energy level, ensuring the delivery matches the seriousness of the legal document.
What is the difference between tts-1 and tts-1-hd?
Both models use the same engine and have the same price. They use the same engine and price. Since pricing is identical, you can choose tts-1-hd for clearer audio without any additional cost, ensuring precise pronunciation for complex legal terminology and contract details.
Do voices support all 23 languages equally?
Yes, every voice can speak any of the 23 supported languages. This includes Arabic, Chinese, English, German, Spanish, and others. If the language is not set, it is guessed from the script, but for non-English Latin text, you should explicitly set the language parameter to ensure accurate pronunciation and formatting.
Can I adjust the tone of a voice?
You cannot adjust emotional tone or add emotion tags. You can only adjust the speed, which ranges from 0.25 to 4.0. This allows you to slow down the delivery for complex legal terms or speed it up for routine reviews. The voice character remains fixed; only the pacing changes.
Are the voices based on real people?
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 voices but preset interpretations of public domain source material.
Is the audio clear in noisy environments?
The API returns audio files in formats like mp3, wav, opus, flac, or pcm. You can choose high-quality formats like wav or flac for better clarity. While the API does not guarantee playback conditions, using a higher fidelity format ensures the best possible sound quality for listening in various office environments.
Do voices sound robotic?
The voices are based on licensed human recordings from LibriVox, so they sound natural rather than robotic. They are synthetic but designed for realistic narration. Each file includes an inaudible watermark to identify it as synthetic speech, but the auditory experience is intended to be clear and human-like for legal reviews.
Can I mix voices in one document?
The API processes one voice per request. To mix voices, you must split your document into parts, generate audio for each part with a different voice, and join the files. You can split texts into parts of up to 4096 characters and combine the resulting audio files manually.
How does speed affect clarity?
Speed ranges from 0.25 to 4.0. Slower speeds (e.g., 0.5) improve clarity for complex legal terms, while faster speeds (e.g., 2.0) help process large volumes of text efficiently. Typical generation is roughly real-time, so adjusting speed helps balance detailed review with rapid scanning of case files.
Is there a best voice for Spanish?
All eight voices can speak all 23 supported languages, including Spanish. There is no single best voice; each has a distinct character. Aria is a warm narrator, while Atlas is neutral. You can choose any voice based on your team's preference for tone, as all support Spanish equally well.
Do voices handle legal numbers well?
The API reads text input directly. It does not have a custom pronunciation lexicon, so numbers are read according to the voice's standard training. For specific formatting, ensure your text is clearly written. The system supports up to 4096 characters per request, allowing you to manage long numerical sequences in contracts.
Can I hear a sample of each voice?
The /v1/voices endpoint provides information about voices, languages, formats, and price. However, the API itself returns audio files via the speech endpoint. You can generate a short sample text for each voice to hear the difference.
Neutral Voices for Facts
Legal teams require precision when processing case files and contracts. Neutral voices deliver facts with clear articulation, ensuring that dense terminology is understood without distraction. These presets prioritize readability and steady pacing over emotional inflection, making them ideal for long-form document review.
Atlas: News Style
Atlas offers a neutral, news-style delivery that handles large volumes of text efficiently. Its steady rhythm helps legal teams parse complex sentences without fatigue. This voice is particularly effective for reading aloud lengthy briefs where consistent tone aids retention.
- Ideal for formal briefs and case summaries
- Maintains steady pace during long reviews
- Neutral tone reduces listener distraction
Sage: Instructional
Sage provides a calm, instructional tone suitable for internal training materials or procedural guides. Its clarity ensures that step-by-step legal procedures are conveyed accurately. This voice works well when the content requires careful attention to detail rather than narrative engagement.
- Best for procedural documents
- Clear enunciation for technical terms
- Supports focused internal review sessions
Use Cases
These voices are optimal for processing standard legal documents where consistency matters more than expressiveness. Teams can use Atlas for external briefs and Sage for internal policy manuals. Both integrate seamlessly with the API, allowing bulk processing of contract clauses.
Narrative Voices for Context
For legal teams reviewing complex case histories, narrative voices provide the necessary context to maintain focus during long document reviews. These voices provide clear audio for legal documents, helping teams process text without losing track of the narrative flow.
Aria: Warm Narrator
Aria offers a warm, clear narrator style that makes dense case files easier to digest. Her tone is professional yet approachable, reducing listener fatigue during extended review sessions of lengthy contracts or discovery documents.
- Ideal for reading lengthy case summaries
- Reduces fatigue during long reviews
- Clear enunciation for complex legal terms
Luna: Storytelling
Luna brings a soft, storytelling quality to legal narratives. She is particularly effective for presenting chronologies of events or witness statements where the sequence of actions matters. Her gentle pacing helps listeners follow the logical progression of arguments.
- Best for chronological event summaries
- Gentle pacing for complex arguments
- Enhances retention of witness statements
Use Cases
These voices are best suited for internal training modules, client briefing summaries, and accessibility reviews where the emotional context of the legal matter needs to be conveyed clearly without the distraction of a monotone delivery.

Energetic Voices for Updates
Energetic voices excel when legal teams need to broadcast urgent updates or highlight key changes in complex documents. For internal team communications, these voices maintain high engagement levels during lengthy audio reviews.
Using the right tone ensures that critical deadlines and procedural shifts are heard clearly, reducing the risk of missed details during rapid information dissemination across large teams.
Felix: Promotional
Felix delivers an energetic, promo-style delivery that cuts through background noise effectively. This voice is ideal for highlighting specific clauses or urgent notices within large case files.
Its dynamic pacing keeps listeners engaged during repetitive review sessions, making it suitable for summarizing contract amendments or highlighting new regulatory requirements.
- Best for urgent deadline notifications
- Ideal for contract amendment summaries
- Effective in noisy office environments
Nova: Upbeat
Nova offers a bright, upbeat tone that transforms dry legal text into approachable audio. This voice is particularly effective for training materials or onboarding new team members.
Its clear enunciation ensures that complex legal terminology is processed accurately, supporting both novice and experienced reviewers in understanding nuanced arguments.
- Perfect for training modules
- Great for onboarding new associates
- Clear enunciation for complex terms
Use Cases
Apply Felix for high-priority alerts regarding filing deadlines or court date changes. Use Nova for educational content explaining new legal procedures or software updates.
Both voices support the 23 available languages, allowing consistent audio generation for domestic or international legal partners.
- Filing deadline alerts
- Court date change notifications
- Procedural training videos
- Software update announcements
Conversational Voices for Meetings
Conversational voices excel when legal teams need to review meeting transcripts or internal communications that mimic natural dialogue. These presets bridge the gap between formal reading and casual listening, reducing cognitive fatigue during long review sessions.
Choosing the right tone ensures that the nuance of spoken agreements or casual updates is preserved without sounding robotic or overly dramatic.
The API supports both male presets, allowing teams to standardize their audio workflow while maintaining distinct auditory cues for different types of content.
Milo: Friendly
Milo delivers a warm, conversational tone ideal for internal emails or informal meeting notes. His natural inflection helps teams process casual updates quickly without losing the friendly context of the original message.
This voice works best for text that feels like a direct conversation, such as follow-up emails or team sync summaries.
- Best for: Internal emails, meeting recaps, casual updates
Orion: Deep
Orion provides a deep, authoritative sound that lends gravity to recorded discussions. His steady pace and clear enunciation make him suitable for summarizing key points from lengthy deliberations or client calls.
Use Orion when you need the audio to convey seriousness without sounding like a news broadcast.
- Best for: Client calls, deliberation summaries, serious discussions
Use Cases
Apply Milo for reviewing informal Slack-style transcripts or quick internal memos where a human touch aids comprehension. Use Orion for recording formal meeting outcomes or client communication logs where clarity and tone depth matter.
Both voices support the 23 available languages, allowing accurate pronunciation for international team communications.
Voice Comparison Table
Selecting the right voice requires balancing tone with technical precision. This comparison helps legal teams evaluate how each preset handles dense case files, ensuring clarity without unnecessary emotional inflection.
Tone Analysis
Aria and Luna offer warm, narrative tones suitable for contextual briefs. Atlas and Sage provide neutral, clear delivery ideal for factual reporting. Felix and Nova bring energetic energy to updates, while Milo and Orion support conversational meeting notes.
Speed Range
All voices support speeds from 0.25x to 4.0x. Adjusting speed allows reviewers to parse complex legal jargon more quickly or listen closely to nuanced clauses. Standard legal reading speed is approximately 150–160 words per minute.
Clarity Score
Clarity is maintained across all eight presets. Each voice is designed for precise pronunciation, essential for legal terms. The neutral voices like Atlas and Sage are well-suited for dense text, and all are optimized for legibility.
Selecting by Document Type
Different legal documents require distinct vocal characteristics to ensure clarity and appropriate tone during playback. Selecting the right voice depends on whether you are reviewing dense contractual language, listening to argumentative briefs, or processing routine internal correspondence.
Using a consistent voice for each document type helps legal teams distinguish content quickly, reducing cognitive load during long review sessions and improving accessibility for team members.
Contracts
Contracts are dense with specific terms and conditional clauses. A neutral, instructional voice like Sage provides steady pacing without emotional inflection, allowing attorneys to focus on precise wording. Alternatively, Atlas offers a clear, news-style delivery that handles long paragraphs efficiently without introducing interpretive bias.
- Use Sage for detailed clause review
- Use Atlas for standard contract reading
Briefs
Legal briefs often contain complex arguments and narrative elements. A warm narrator like Aria can help convey the context of case facts, while a deep, documentary voice like Orion adds weight to persuasive arguments without sounding overly dramatic. These voices help maintain engagement during lengthy textual reviews.
- Use Aria for case background context
- Use Orion for persuasive argument emphasis
Emails
Internal legal emails are typically shorter and conversational. Milo’s friendly, conversational tone makes quick updates and queries feel natural and less rigid. For more formal external correspondence, Atlas remains a safe choice for maintaining professional neutrality in recorded communications.
- Use Milo for casual internal updates
- Use Atlas for formal external notes
Technical Specifications
To ensure seamless integration with your legal review workflows, the API supports five standard audio formats. You can output files as MP3, WAV, OPUS, FLAC, or PCM, depending on your storage and playback requirements.
Input text is processed with a limit of 4,096 characters per request. For lengthy contracts, split the text into chunks and join the resulting audio files to maintain precise pronunciation and formatting throughout the document.
Format Support
Choose from MP3 for general playback, WAV for high-fidelity editing, or FLAC for lossless archival storage. PCM and OPUS offer additional flexibility for specific software integrations used by large legal teams.
Each format preserves the synthetic watermark, ensuring you can verify the audio as generated by LexiSpeak without affecting the listening experience.
- MP3: Standard compressed audio for quick reviews
- WAV: Uncompressed lossless audio for editing
- FLAC: Lossless compression for archival storage
- PCM: Raw audio data for specialized systems
- OPUS: Efficient compression for web playback
Language Fit
The API supports 23 languages, including English, Spanish, German, and French. Any voice can speak any supported language, allowing you to standardize on a single voice model across diverse case files.
Set the language parameter explicitly for non-English Latin-script text to ensure accurate pronunciation. If omitted, the system guesses the language, defaulting to English for Latin scripts.
- Supports 23 languages including Arabic, Chinese, and Japanese
- All voices support all supported languages
- Explicit language setting recommended for non-English texts
Audio Quality
Generation speed is roughly real-time, processing approximately 25 characters of text per second. Short phrases may take a few seconds, while longer texts are processed sequentially, potentially queuing under high load.
The audio quality is consistent across all voices, with no artificial emotion tags or SSML support. Focus on clarity and precise pronunciation rather than expressive variations for legal documents.
- Real-time processing speed for long texts
- Sequential processing ensures consistent output
- No SSML or emotion tags for precise control

Trial Recommendations
Before committing to a voice for your entire case file, start with a short, representative excerpt. Generate a short sample to verify the synthesis engine handles your specific legal terminology correctly. This quick check helps you gauge the base speed and audio quality without exhausting your trial credit.
Use this test phase to ensure the voice renders complex abbreviations and names accurately. Since the API guesses language from script, explicitly setting the language code for non-English documents prevents mispronunciation. Review the output for natural pacing, especially in dense paragraphs typical of contracts.
Test with Short Text
Generate a 50-character sample to verify synthesis handles specific legal terms. This verifies base speed and quality without wasting credit.
Explicitly set the language code for non-English documents to prevent script-based guessing errors. Review output for natural pacing in dense paragraphs.
- Use standard contract clauses as test data.
- Verify handling of abbreviations like 'v.' or '§'.
- Check pacing for dense legal jargon.
Check Pronunciation
Listen for accurate rendering of proper nouns, Latin terms, and statutory references. The voices are trained on licensed recordings, but specific legal jargon may require manual adjustment of the input text if mispronounced.
Test names of parties and case citations. Ensure the voice does not stumble over numbers or dates, which are frequent in legal documents.
- Test case names and citations.
- Verify Latin term pronunciation.
- Check number and date formatting.
Evaluate Clarity
Assess the voice’s ability to maintain clarity over longer passages. Use the speed parameter (0.25–4.0) to find the optimal listening pace for your team. Faster speeds may reduce clarity for complex arguments.
Ensure the audio is distinct from background noise. The inaudible watermark marks the file as synthetic, which is useful for internal tracking but does not affect intelligibility.
- Adjust speed for optimal listening.
- Ensure audio is distinct from noise.
- Verify synthetic watermark does not affect clarity.
Team Preferences
Selecting a voice depends on whether you are reviewing dense case files alone or presenting updates to a group. Solo reviewers often prefer neutral tones that reduce listening fatigue during long sessions, while team settings benefit from voices that convey clarity and engagement for multiple listeners.
Client sharing requires a professional tone that sounds authoritative yet approachable. The chosen voice should remain consistent across different document types to maintain a unified auditory identity for your firm’s internal and external communications.
Solo Review
For individual review of lengthy contracts, use Atlas or Sage. These voices provide a steady, neutral pace that helps maintain focus during detailed clause analysis without causing ear fatigue.
Set the speed to 1.0x for optimal comprehension. This ensures every legal term is pronounced clearly, allowing you to catch subtle nuances in the text.
- Use Atlas for factual summaries.
- Use Sage for instructional guidelines.
Team Listening
When sharing briefs with colleagues, Nova or Felix add a touch of energy that keeps the group engaged. These voices are bright and clear, making them suitable for team meetings or shared listening sessions.
Ensure all team members have the same voice profile configured in their API keys to maintain consistency across shared audio files.
- Use Nova for upbeat updates.
- Use Felix for energetic promos.
Client Sharing
For external presentations, Orion or Aria offer a polished, professional sound. Orion’s deep tone conveys authority, while Aria’s warmth makes complex legal concepts more accessible to non-experts.
Test the audio with a sample contract before sending to ensure the tone matches your firm’s brand image.
- Use Orion for deep authority.
- Use Aria for warm clarity.
Licensing Details
LexiSpeak’s voice models are built on licensed recordings from the LibriTTS-R dataset, which derives from LibriVox public-domain audiobook readings. This foundation ensures consistent audio quality while respecting the original creators’ rights. We do not offer custom voice cloning, so your team receives the same reliable presets used by all subscribers.
LibriTTS-R Source
The underlying voice data originates from the LibriTTS-R dataset, sourced from LibriVox’s collection of public-domain audiobooks. These recordings provide a diverse range of natural speech patterns, ensuring that voices like Atlas or Luna sound professional and clear. This origin supports our commitment to transparency regarding the training data used for synthesis.
- Source: LibriTTS-R dataset
- Origin: LibriVox public-domain audiobooks
- Type: Licensed recordings for API use
CC BY 4.0
The LibriVox content is released under the Creative Commons Attribution 4.0 International License. This allows us to legally incorporate these recordings into our TTS engine. We adhere to the attribution requirements, ensuring that the original source material is credited appropriately in our documentation. This license supports our model of providing speech without complex enterprise restrictions.
- License: CC BY 4.0
- Rights: Commercial use allowed
- Attribution: Required in documentation
Public Domain
Many of the source audiobooks are in the public domain, meaning their copyright has expired or was never claimed. This reduces legal risk and ensures stable availability of the voice models. You can rely on consistent voice characteristics across your legal documents, knowing the source material remains freely accessible and stable for long-term use.
- Status: Public domain where applicable
- Stability: High availability
- Risk: Minimized copyright issues
Watermark Information
Every audio file generated by LexiSpeak contains an inaudible digital watermark. This mark is embedded directly into the audio waveform and remains imperceptible to the human ear during standard playback. It serves as a permanent identifier that the content was produced by our synthesis engine rather than a human speaker.
This feature ensures transparency for legal teams reviewing large volumes of converted case files or contracts. By marking each file, the watermark provides a clear audit trail, distinguishing synthetic speech from original human recordings without affecting the listening experience or document integrity.
Inaudible Mark
The watermark is woven into the audio data at a frequency level that does not alter the perceived tone, speed, or clarity of the voice. Whether you are using the neutral Atlas voice for news or the warm Aria voice for narrations, the mark remains consistent and undetectable during standard playback.
This ensures that the audio remains suitable for both internal review and accessibility purposes without any audible artifacts or changes in quality.
Synthetic ID
Each generated file carries an inaudible watermark that marks it as synthetic speech.
The identifier is designed to be robust against common audio processing operations, ensuring it survives compression and format conversions used in typical legal document workflows.
Detection
While the mark is inaudible to humans, specialized software can detect the presence of the synthetic watermark. This allows legal teams to verify that a recording is AI-generated, which is useful for compliance and transparency requirements.
The detection process does not require downloading the file or accessing the API, making it easy to confirm the synthetic nature of the audio during review.
This feature supports your need for precise documentation and clear attribution in large-scale legal text processing.

Final Selection Criteria
Choosing the right voice for legal TTS prioritizes clear articulation over emotional nuance, ensuring complex case details remain intelligible during review. Consistency in tone prevents listener fatigue during long document sessions, maintaining focus on critical legal arguments without distraction from variable delivery styles.
Requests are processed one after another, so under load a request may wait in a queue.
Clarity Over Emotion
Legal documents require precise pronunciation of technical terms and names rather than expressive delivery. Selecting neutral or instructional voices like Atlas or Sage ensures that every syllable is distinct, reducing ambiguity in critical clauses or jury instructions where misinterpretation could impact case outcomes.
- Prioritize voices with high articulation scores for dense text.
- Avoid emotional inflections that may obscure legal nuances.
- Test with domain-specific vocabulary for accuracy.
Consistency
Maintaining a uniform voice across all generated audio files helps teams establish a familiar listening rhythm. This consistency is crucial for accessibility and long-form document reviews, where varying tones can disrupt concentration and reduce comprehension rates over extended listening periods.
- Use the same voice for all documents in a case.
- Standardize speed settings across team members.
- Ensure uniform format output for easy integration.
Load Testing
Before full deployment, process a representative sample of your heaviest case files through the API to verify performance. Monitor response times and queue behavior to ensure that concurrent requests from multiple team members do not cause unacceptable delays in audio generation.
- Simulate peak usage with 10+ simultaneous requests.
- Measure average generation time per 1,000 characters.
- Verify rate limit handling (60 req/min).
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. These comprehensive legal documents streamline compliance for our large distributed teams across multiple jurisdictions.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- openai-node on GitHub
- Node.js documentation
- LibriTTS-R paper (arXiv 2305.18802)
- W3C WAI: making audio and video accessible
- openai-python on GitHub
- TRON developer hub: TRC-20
- curl documentation
- OpenAI API reference: create speech
- Python documentation
- LibriTTS-R corpus (OpenSLR 141)
- LibriVox public-domain audiobooks
