Frequently asked questions
Is the input text stored after audio generation?
Input text is used only to produce the audio and is not stored. Only request metadata, including time, characters, and status, is retained for billing purposes. This ensures that the specific content of your case files remains transient during processing.
Are the voices licensed for commercial legal use?
Voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. This foundation supports commercial use within your legal workflows without additional licensing complexities for the voice assets themselves.
Does the API use my data to train models?
No. Input text is used exclusively to generate the audio file and is not used for training models. Your confidential legal documents remain isolated from your data used for model improvement.
Can I identify that the audio is synthetic?
Yes. Every generated audio file carries an inaudible watermark that explicitly marks it as synthetic speech. This allows for easy verification that the audio was generated by our API rather than recorded by a human.
Is there a guarantee of audio quality for legal terms?
We do not offer an SLA or uptime guarantee. While voices are clear and precise, audio quality depends on network conditions and queue status. We do not guarantee specific pronunciation accuracy for highly specialized legal jargon.
What happens to my data if I delete my account?
Request metadata (time, characters, status) is kept for billing. Input text and generated audio are not stored.
Are the voices based on real people?
The voices are based on LibriTTS-R (CC BY 4.0) and LibriVox public-domain recordings. This generally mitigates copyright risks for the voice synthesis itself. However, you remain responsible for the copyright status of the input text you provide.
Is there a risk of copyright infringement with the voices?
Yes. The API supports standard formats like mp3, wav, opus, flac, and pcm. You can generate high-quality audio for briefs or case files. The synthetic watermark allows you to disclose that the audio is AI-generated if required.
Can I use this for client-facing legal documents?
Input text is not stored after generation, reducing long-term exposure. Data is transmitted via standard HTTPS. However, we do not provide specific certifications like SOC 2 or HIPAA.
Is the API secure for confidential case files?
The voices are based on LibriTTS-R (CC BY 4.0), which typically requires attribution. You should verify the specific CC BY 4.0 requirements for your use case. The API itself does not enforce attribution in the audio file.
Do I need to attribute the voices in my output?
Each request accepts up to 4096 characters. For longer documents, you must split the text into parts and join the resulting audio files. This allows you to process lengthy case files by breaking them into manageable chunks.
Is there a limit to how long the audio can be?
There are no restrictions on commercial use. You can use the API for any purpose, including client-facing documents, internal briefs, or training materials. The only limits are rate limits (60 requests per minute) and character counts per request.
Are there any restrictions on commercial use?
Every generated file contains an inaudible watermark marking it as synthetic speech. You can use a watermark detection tool to verify that the audio was produced by our API and has not been altered or replaced with human recording.
Data Privacy and Retention
VerdictRead ensures that input text is used exclusively for the purpose of generating audio and is not retained for model training. This isolation prevents your case files or briefs from influencing future engine improvements, maintaining strict confidentiality for sensitive legal data.
Input text not used for training
Unlike many consumer-grade TTS services, VerdictRead does not use your submitted text to refine its neural networks. Your documents remain isolated to the generation process, ensuring that proprietary legal language or unpublished case details are not absorbed into the general training dataset.
No storage of generated audio
Generated audio files are delivered directly in the API response and are not stored on our servers after transmission. This minimizes the attack surface and ensures that your converted legal documents do not persist in our cloud infrastructure beyond the immediate delivery window.
Metadata kept for billing only
We retain request metadata, including timestamps, character counts, and status codes, solely for accurate billing and usage verification. This metadata does not include the content of the text or the audio itself, preserving the integrity of your data while ensuring transparent invoicing.
Voice Licensing for Legal Use
Legal teams require clear provenance for all generated assets. Our voices are derived from the LibriTTS-R dataset, ensuring that the underlying audio source material is properly licensed for commercial and institutional use. This foundation supports the integrity of your legal document recordings.
LibriTTS-R licensed recordings
The eight available voice presets are synthesized from the LibriTTS-R dataset. Each voice is a distinct preset, providing consistent tone for briefs and case files.
- Voices are based on licensed recordings from the LibriTTS-R dataset.
LibriVox public domain source
The source material originates from LibriVox, a collection of public domain audiobook readings. By utilizing these existing recordings, we ensure that the base audio data is free from restrictive copyright claims that often complicate third-party voice synthesis. This allows for broad usage within legal departments without additional licensing fees.
- Source material comes from LibriVox public-domain audiobook readings.
CC BY 4.0 license terms
The LibriTTS-R dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. This standard license permits commercial use and modification, provided appropriate credit is given. For internal legal audio generation, this license structure supports compliant usage without requiring complex legal review of each audio file.
- Dataset is released under CC BY 4.0 license terms.
Synthetic Audio Identification
Every audio file generated by VerdictRead contains an inaudible watermark that marks it as synthetic speech. This ensures immediate identification of AI-generated content without affecting the clarity required for legal reviews.
The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0), from LibriVox public-domain audiobook readings. They maintain a consistent, synthetic quality that distinguishes them from human narrators, preventing misattribution in archival records.
Inaudible watermark in audio
Each synthesized file carries a digital watermark embedded within the audio track. This technical marker confirms the content is machine-generated, providing an audit trail for compliance teams verifying document provenance.
- Embedded in every generated file
- Transparent to standard audio players
- Verifiable for legal authenticity
Clear distinction from human voice
The synthetic nature is evident in the consistent pitch and lack of human breath or hesitation. This clarity helps legal teams distinguish between original recordings and converted text files during evidence review.
No celebrity voice imitation
VerdictRead uses eight standard presets (e.g., Atlas, Sage) derived from public-domain audiobook readings. We do not clone specific individuals or mimic celebrities, ensuring no unauthorized likeness rights issues arise in commercial or public filings.
Accuracy for Legal Terminology
Legal departments require precise articulation of complex terminology. Our text-to-speech engine delivers consistent pronunciation across all 23 supported languages without requiring manual intervention. The system handles standard legal phrasing reliably, ensuring that case files and briefs are rendered clearly for review.
The absence of a custom pronunciation lexicon means no additional configuration is needed for standard legal terms. While specialized jargon follows default patterns, the eight voice presets provide uniform clarity. This reduces setup time for high-volume processing of large document sets.
Consistent pronunciation
The API ensures stable articulation across requests. Each character is processed uniformly, maintaining clarity for dense legal text. This consistency supports accurate review by legal teams without variability between generations.
No custom lexicon
There is no custom lexicon for specialized terms. The system relies on standard phonetic rules inherent to the eight voice presets. This simplifies integration, as no manual entry of domain-specific pronunciations is required before processing.
Teams can proceed directly to API integration without defining pronunciation maps. This reduces initial configuration overhead for legal document workflows.
Standard voice presets only
Only the eight predefined voices are available. There is no option for custom voice design or cloning. Users select from neutral, clear presets like Atlas for news-style readings or Sage for instructional content, ensuring uniform audio quality across large volumes of text.
These fixed options eliminate variability in voice characteristics, which is critical for maintaining a consistent audit trail of generated audio files.
- Atlas: Neutral, news-style delivery
- Sage: Calm, instructional tone
- No custom voice cloning available
Limitations for Legal Context
Legal texts often require precise tonal delivery without emotional embellishment. The API uses standard voice presets, meaning you cannot inject specific emotional cues into the audio output. This ensures a consistent, neutral baseline suitable for reading case files, though it lacks the nuance of human narration.
No emotion tags
The system does not support emotion tags. You cannot specify joy, seriousness, or urgency within the request body. The voice remains consistent according to its preset character, which is suitable for straightforward legal readings but may feel monotonous for long briefs.
No SSML
Unlike some competitors, this API does not support Speech Synthesis Markup Language (SSML). You cannot use tags to control pauses, pronunciation, or emphasis. Complex formatting must be handled by splitting the text into separate requests or adjusting the input text manually.
Fixed speed controls only
You can adjust the playback speed between 0.25x and 4.0x using the speed field. However, this is a global setting for the entire request. You cannot vary speed dynamically within a single audio file to highlight specific legal clauses or definitions.
Support for Large Teams
VerdictRead supports large legal teams through a single active API key per account. This simplifies access management for distributed departments, ensuring consistent usage tracking across all team members without complex licensing structures.
Support is provided via email only, with no phone number or live chat available. There is no enterprise contract, SLA, or uptime guarantee.
Single API key management
Each account maintains one active key, which can be regenerated at any time. When a new key is issued, the previous one stops immediately, providing clear audit trails for large teams requiring strict access control without managing multiple credentials.
- Single active key per account
- Immediate revocation upon new key issuance
- No seat or minimum spend requirements
Email-based support only
Technical assistance is handled exclusively through email. There is no live chat, phone support, or dedicated account manager. Teams must rely on asynchronous communication for resolving integration issues or billing inquiries, which may involve waiting times during peak usage.
No enterprise SLA
The service does not offer an enterprise contract, SLA, or uptime guarantee. Requests are processed sequentially, meaning under load a request may wait in a queue. Teams should account for this when integrating into time-sensitive legal workflows.

Comparison with Human Narration
The API processes requests sequentially, generating audio at roughly real-time speed (25 characters per second). For large briefs, this eliminates the scheduling delays and studio booking times associated with human narrators, enabling immediate audio delivery upon request submission.
Faster turnaround time
Standard processing handles texts up to 4096 characters per request. While short phrases take a few seconds, longer documents are split and joined automatically. This allows legal teams to receive audited audio files without waiting for human recording schedules.
- Real-time generation speed (~25 chars/sec)
- Automatic splitting for long documents
- No studio booking delays
Lower cost per hour
At $10 per million characters, the cost is significantly lower than hourly human narration rates. Failed requests incur no charge, ensuring you pay only for successful audio production. This model removes minimum session fees common in human narration contracts.
- $0.01 per 1,000 characters
- No charge for failed requests
- No minimum session fees
Consistent quality across volumes
Human narrators may vary in tone or speed over long sessions or across multiple recording days. The API maintains identical voice characteristics, pitch, and articulation for every request, ensuring uniform audio quality whether generating one brief or ten thousand.
- Identical voice output every time
- No vocal fatigue or variation
- Uniform tone for all documents
Technical Requirements for Compliance
To ensure strict compliance with legal data handling protocols, all interactions with the TTS engine occur over secure HTTPS endpoints. This guarantees that sensitive case files and briefs are encrypted in transit during transmission from your internal systems to our cloud API.
Authentication is managed via a single active Bearer token per account, providing a clear audit trail for every request. The standard JSON request format allows seamless integration into existing legal tech stacks without requiring proprietary SDKs or complex middleware adapters.
Secure API endpoints
The API base is located at https://api.playtko7.com/v1. All requests use standard HTTPS, ensuring that input text and generated audio streams are protected from interception during transmission.
Bearer token authentication
Access is secured using a unique Bearer token issued to your account. This token must be included in the Authorization header of every request. A new key invalidates the previous one immediately, allowing precise control over access points.
- Single active key per account
- Immediate invalidation upon key rotation
- Standard HTTP header implementation
Standard JSON request format
Requests follow the standard JSON structure, compatible with common HTTP libraries. This ensures predictable data exchange without requiring custom parsing logic for proprietary formats.
- JSON body structure
- Standard HTTP methods
Voice Selection for Legal Tone
Selecting the correct voice model ensures that complex legal terminology is processed with the necessary clarity and professionalism. The API provides eight distinct presets, allowing legal departments to choose voices that align with specific document types, from briefs to case files.
For formal readings, neutral and instructional tones prevent distraction while maintaining high intelligibility. This section outlines which voices best suit strict corporate environments where consistent, unbiased delivery is required for accurate information retention.
Neutral voices for readings
Atlas offers a neutral, news-style delivery ideal for reading lengthy case files or regulatory updates. His tone is steady and unobtrusive, ensuring that the content remains the focus without emotional interference.
Sage provides a calm, instructional voice suitable for explanatory briefs or instructional legal documents. His steady pace aids comprehension of dense material without introducing subjective tonal variations.
- Atlas: Neutral, news-style delivery for case files.
- Sage: Calm, instructional tone for explanatory briefs.
Clear articulation for details
Aria and Nova deliver clear, crisp articulation essential for precise legal terminology. Aria’s warm, clear narrator style ensures that complex clause structures are distinguished clearly, while Nova’s bright, upbeat tone keeps attention high during rapid readings.
Milo’s friendly, conversational style can be effective for internal memos or less formal legal communications, providing a natural flow without sacrificing clarity.
- Aria: Warm, clear narrator for complex clauses.
- Nova: Bright, upbeat tone for attention retention.
- Milo: Friendly, conversational style for internal memos.
Consistent pitch and speed
All voices maintain consistent pitch and speed, eliminating the variability found in human narration. This uniformity is critical for large-scale document processing where consistency across hundreds of pages is required.
Users can adjust speed between 0.25x and 4.0x to match specific listening preferences or time constraints, ensuring that every document is delivered at an optimal processing rate.
- Uniform pitch across all generated audio.
- Adjustable speed from 0.25x to 4.0x.
- No tonal variation between paragraphs.
Data Security Measures
VerdictRead processes legal texts through a secure API endpoint, ensuring that input data is transmitted via encrypted channels. The system does not share your case files or briefs with any external third-party vendors for processing or analysis, maintaining strict control over your proprietary information.
No third-party data sharing
Our architecture ensures that your text input is used exclusively for immediate audio synthesis. We do not distribute, sell, or license your legal documents to any third-party entities, preserving the confidentiality of your client data and internal communications.
Encryption in transit
All API requests and audio responses are secured using standard TLS encryption. This ensures that your legal texts and generated audio files remain protected during transmission between your systems and our servers, preventing unauthorized interception during the conversion process.
No long-term text storage
Input text is processed solely to generate the audio file and is not retained for training or long-term storage. Once the audio is delivered, the original text content is not archived, minimizing the risk of data exposure from stored records.
Auditability of Audio Generation
Every request generates a verifiable audit trail, ensuring that legal teams can trace exactly which text was converted. This metadata supports compliance reviews by documenting the conversion process from input to output.
Request metadata logging
The system retains request metadata including time, characters processed, and status for billing and audit purposes. This data remains available for internal reviews without storing the actual text or generated audio files.
- Logs request time, character count, and status
Character count verification
Each request is charged based on precise character counts, including spaces and punctuation. This ensures accurate cost tracking and provides a definitive record of the volume of text processed for legal documents.
Timestamp records
Timestamps are recorded for every API call, providing a chronological record of when each document section was converted. This allows legal departments to reconstruct the exact sequence of audio generation events.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence.
23 languages
Any voice speaks any of these languages.
