Frequently asked questions
What is the difference between TTS and human voice over?
Our API generates synthetic speech from text using licensed recordings, eliminating manual studio booking and scheduling. Unlike human voice overs, it offers instant scalability for high-volume localization projects without the need for coordinating large teams or physical recording sessions.
How accurate is the language detection?
The API guesses the language from the script if not specified, defaulting to English for Latin text. For non-English Latin-script text, you should explicitly set the language parameter to ensure accuracy. It supports 23 specific languages, excluding others like Vietnamese or Thai.
Can TTS match the lip movements of a video?
No. The API generates audio files from text without visual synchronization data. It does not offer lip-syncing features or video editing capabilities. You must handle audio-video synchronization manually in your post-production workflow after receiving the generated audio files.
What is the latency of the API?
Generation takes roughly real time, about 25 characters per second, with short phrases taking a few seconds. Requests are processed sequentially, so under load, a request may wait in a queue. This latency is suitable for batch processing, not live streaming.
Do I need to know how to code to use the API?
You need basic coding skills to send POST requests to the endpoint. The API uses a standard JSON body compatible with the OpenAI SDKs. You can use the provided curl examples or integrate directly into your localization pipeline using any HTTP client.
How are voice licenses handled?
Voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobook readings. Each generated file carries an inaudible watermark marking it as synthetic speech. You do not need to manage individual voice licenses for each generated file.
Is the audio quality suitable for broadcast?
The API provides synthetic speech using the tts-1 or tts-1-hd models. While suitable for many professional use cases, it does not guarantee broadcast-specific standards or offer specific audio mastering. It is designed for efficient localization workflows rather than traditional broadcast master files.
Can I change the speed of the voice?
Yes. You can adjust the speed parameter between 0.25 and 4.0 in your request. This allows you to speed up or slow down the generated audio to match specific timing requirements for your video localization projects without re-generating the entire file.
What happens if the API is down?
We do not provide an SLA or uptime guarantee. If the service is unavailable, requests may fail. Failed requests are free, and the charge is returned in full. We support customers via email only, so check your inbox for status updates.
Is the API suitable for live dubbing?
No. The API processes requests sequentially with a typical speed of roughly 25 characters per second. It does not support streaming or WebSocket connections. It is optimized for batch processing high volumes of voice-overs.
How do I handle special characters in text?
Send the text directly in the JSON body. Spaces and punctuation count as characters toward the 4096-character limit. The API handles standard text input; there are no special SSML tags or phoneme dictionaries required for basic character handling.
Can I use the API for multiple languages simultaneously?
Yes. You can specify the language parameter for each request, supporting 23 languages including Arabic, Chinese, and Swahili. Each voice preset can speak any of these languages, allowing you to generate audio for diverse localization projects in a single integration.
What is the maximum file size for audio output?
The API does not specify a maximum file size, but limits input text to 4096 characters per request. For longer texts, you split the input and join the resulting audio files. File size depends on the format (mp3, wav, etc.) and duration.
Do I need a dedicated server to use the API?
No. The API is cloud-based. You send requests to the endpoint from any server or device with internet access. We handle the infrastructure, so you do not need to manage servers, storage, or processing power for audio generation.
Definition of TTS API
SyncVoice is a cloud-based text-to-speech API designed for localization agencies. It converts written text into audio files via HTTP requests, enabling programmatic generation at scale rather than manual recording.
Unlike consumer interfaces, our API provides direct JSON-based control over voice selection, speed, and format. This allows integration into existing localization pipelines without manual intervention.
Text to Speech Technology
The API synthesizes speech using neural models trained on licensed recordings. Users send text input and receive audio files in formats like MP3, WAV, or FLAC.
API vs User Interface
While user interfaces offer manual playback, the API allows bulk processing. Agencies can generate thousands of voice-overs automatically using standard HTTP requests.
- No graphical interface required for production
- Direct JSON payload integration
- Supports automated workflows
Programmatic Control
Developers control parameters like voice, speed, and language via code. The endpoint accepts standard JSON bodies, compatible with the OpenAI SDK structure for easy adoption.
- Adjustable speed (0.25–4.0)
- Specific voice preset selection
- Language code specification
Role in Localization
SyncVoice API replaces traditional human recording workflows by generating audio programmatically. Localization agencies submit text via API, receiving synthesized speech without scheduling voice talent or booking studio time. This shifts production from manual coordination to automated, high-volume generation.
Replacing Human Recording
Agencies eliminate the logistical overhead of hiring narrators for each language. Text is sent directly to the engine, which returns audio files ready for insertion into video timelines. This removes dependency on individual talent availability and allows immediate scaling.
- Eliminates studio booking and talent scheduling
- Direct text-to-audio pipeline
- No manual coordination required
Speed of Delivery
Generation occurs in near real-time, typically processing roughly 25 characters of text per second. Longer texts are split into manageable chunks and stitched automatically. This enables rapid turnaround for large dubbing projects without waiting for human narration cycles.
- Processing at ~25 chars/sec
- Automatic chunking for long texts
- Immediate audio file return
Consistency Across Projects
Using fixed voice presets ensures uniform audio quality across all localized versions. The same voice model generates all output, maintaining tonal consistency regardless of project size. This uniformity simplifies quality control for multi-language video releases.
- Uniform voice tone across languages
- Standardized audio quality
- Simplified quality assurance
Technical Requirements
To operate the SyncVoice API, your infrastructure requires stable internet connectivity to submit JSON payloads and receive audio files. Requests are processed sequentially, meaning high-volume batches may queue under load, so ensure your pipeline handles potential delays gracefully without duplicating submissions.
Internet Connectivity
Stable connectivity is essential for submitting JSON payloads and receiving generated audio. Since requests are processed sequentially, high-volume batches may queue during peak times. Your integration should account for these potential delays to avoid duplicate submissions or timeouts.
JSON Payload Structure
Send a POST request to https://api.playbd2.com/v1/audio/speech with a standard JSON body. Required fields include model ("tts-1" or "tts-1-hd"), input (text, max 4096 characters), voice, and response_format. The language field is optional; if omitted, the system guesses the language from the script, defaulting to English for Latin text.
- Model: "tts-1" or "tts-1-hd"
- Input: Max 4096 characters
- Voice: One of 8 presets
- Format: mp3, wav, opus, flac, or pcm
Audio Format Output
The API returns audio files in mp3, wav, opus, flac, or pcm formats. Each file is generated within the response. For texts exceeding 4096 characters, your system must split the input, submit multiple requests, and join the resulting audio files. Failed requests incur no charge and return no audio data.
Integration with Video Tools
SyncVoice API enables localization agencies to replace manual recording workflows with programmatic audio generation at scale. By sending text via JSON payload, teams receive audio files that integrate directly into video post-production pipelines, eliminating the bottleneck of scheduling voice talent for every language version.
Syncing with Video Tracks
The API returns standard audio formats (mp3, wav, opus) that align precisely with video timelines. Teams can match audio duration to visual cuts by splitting input text into parts under 4096 characters, ensuring seamless lip-sync or voice-over alignment without manual trimming.
- Standard formats (mp3, wav, opus, flac, pcm) for direct import into editing software.
- Consistent timing allows predictable alignment with video tracks.
- Generation speed is roughly real time (about 25 characters per second), though requests may wait in a queue under load.
Automated Workflow Pipelines
Integration is achieved by changing the OpenAI SDK base_url to https://api.playbd2.com/v1. Scripts can automate the entire process: submitting translation text, receiving audio, and uploading assets to content management systems or video hosting platforms without human intervention.
- Compatible with existing OpenAI SDKs via simple base_url configuration.
- Batch processing handles multiple language versions simultaneously.
- Automated billing tracks characters used per video asset.
Metadata Tagging
While the API does not embed video metadata, request metadata (time, characters, status) is retained for billing and audit purposes. Teams can tag generated audio files with source text IDs, voice presets, and language codes to maintain organized asset libraries for large-scale dubbing projects.
- Request metadata includes timestamp and character count for auditing.
- Audio files can be externally tagged with voice and language data.
- No in-band metadata is embedded in the audio stream itself.
Quality Considerations
SyncVoice API generates audio using licensed recordings from LibriTTS-R and LibriVox, ensuring a baseline of natural human speech patterns. Each file contains an inaudible watermark identifying it as synthetic, aiding in content verification for localization agencies. The engine processes text up to 4096 characters, delivering consistent output across high-volume projects without the variability of individual human voice actors.
Prosody and rhythm are handled by the underlying model, which respects language-specific nuances. For non-English Latin text, specifying the ISO language code prevents default English inflection, ensuring accurate pacing. This programmable control allows teams to maintain consistent audio quality across diverse dubbing pipelines, replacing manual recording workflows with reliable, scalable generation.
Voice Naturalness
The eight voice presets—such as Aria for narration or Felix for promos—are based on licensed recordings. This foundation provides a realistic baseline that avoids the robotic tone common in older TTS systems. Every generated file carries an inaudible watermark, marking it as synthetic speech for compliance and audit purposes in professional localization workflows.
- Based on LibriTTS-R (CC BY 4.0) and LibriVox public-domain readings
- Eight distinct voice presets for different content types
- Inaudible watermark in every generated audio file
Prosody and Rhythm
The API respects the linguistic structure of the input text. By setting the language parameter, you ensure correct rhythm for non-English Latin scripts, preventing default English intonation patterns. This precise control over pacing is essential for synchronizing audio with video tracks in dubbing, where timing accuracy is critical.
- Support for 23 languages including Arabic, Chinese, and Japanese
- Optional language code overrides script-based guessing
- Consistent rhythm across high-volume project batches
Emotional Tone
SyncVoice does not offer dynamic emotion tags or custom voice cloning. Emotional tone is determined by the selected voice preset—for example, Nova provides a bright, upbeat tone, while Atlas remains neutral for news. This fixed approach ensures consistency across large teams and projects, though it lacks the nuanced emotional shifts of human performers.
- Fixed emotional character per voice preset
- No dynamic emotion tags in requests
- Consistent tone across all generated audio files
Use Cases in Dubbing
SyncVoice API enables localization agencies to replace manual recording workflows with programmatic audio generation, ensuring consistent voice quality across high-volume dubbing projects. By utilizing eight distinct voice presets, teams can maintain brand identity while scaling production efficiently.
Documentary Narration
Use the deep, documentary-style voice 'orion' for historical content or 'atlas' for neutral news formats. The API handles long scripts by splitting them into 4096-character segments, joining the audio files seamlessly for continuous playback.
- Deep, authoritative tone suitable for factual content
- Automatic segmentation for long-form scripts
- Consistent delivery across multiple languages
E-Learning Modules
Generate clear, instructional audio for training materials using voices like 'sage' or 'nova'. With support for 23 languages, agencies can localize educational content rapidly without hiring separate voice actors for each market.
- Clear enunciation for instructional clarity
- Support for 23 languages
- Rapid turnaround for course updates
Promotional Videos
Leverage energetic voices like 'felix' or 'aria' to create engaging promotional content. The API’s speed control (0.25–4.0) allows fine-tuning of pacing to match video visuals precisely, ensuring tight synchronization.
- Adjustable speed for visual sync
- Energetic tones for marketing
- Immediate audio generation
Comparison with Traditional Methods
Traditional localization relies on hiring voice actors for each project, incurring variable per-minute rates and scheduling delays. Our API replaces this with programmatic generation at a fixed cost of $0.01 per 1,000 characters, regardless of volume. This eliminates the premium for urgency or complex casting requirements, providing predictable unit economics for high-volume dubbing pipelines.
Audio delivery shifts from days of studio booking and editing to seconds of API response time. For long-form content, text is split into 4,096-character segments and synthesized sequentially, yielding roughly real-time audio duration. This compression of the production timeline allows localization teams to iterate on scripts without waiting for external talent availability.
Cost Comparison
Manual recording incurs costs per minute of final audio, plus editing fees. The API charges strictly by character count ($10 per million characters), decoupling price from audio length or complexity. Failed requests are free, removing financial risk from technical glitches.
No minimum spend or monthly subscriptions exist. Teams pay only for what they generate, with balances lasting indefinitely. This model favors projects with fluctuating volumes, unlike fixed retainer contracts with studios.
- Fixed $0.01 per 1,000 characters
- No per-minute premium
- Pay-as-you-go, no subscriptions
Time Comparison
Traditional workflows require casting, booking, recording, and post-production, often taking days per language. The API delivers audio in seconds, with typical generation speed matching real-time for long texts. Requests are processed one after another, so under load a request may wait in a queue.
Parallel processing is limited by the single-key rate limit of 60 requests per minute. For massive projects, this linear throughput is predictable, whereas human studio availability is variable and subject to holidays or illness.
- Seconds vs. days for delivery
- Predictable real-time generation speed
- Immediate turnaround for revisions
Scalability Comparison
Human studios scale by adding resources, which increases coordination overhead. The API handles requests sequentially with a rate limit of 60 requests per minute per key. This allows localization agencies to quote fixed rates for diverse language packs without managing complex vendor networks.
Consistency is maintained across all voices and languages using the same engine. Unlike human actors whose performance varies by session, the API delivers identical voice characteristics for every character, ensuring uniform brand tone across entire video catalogs.
- Instant scaling to any volume
- Uniform voice consistency
- No vendor coordination overhead
Limitations of TTS
SyncVoice API delivers consistent, programmatic audio generation, but it operates within defined technical boundaries. Understanding these constraints ensures realistic expectations for high-volume localization workflows where precise control over output is critical.
Lack of Human Nuance
The system generates synthetic speech with a consistent prosody and rhythm. While voices are based on licensed recordings, the output lacks the subtle, unpredictable inflections of human performers. This uniformity is advantageous for scalability but may feel less organic in artistic contexts.
- Consistent intonation across all languages and voices.
- No spontaneous pauses or breath variations.
- Predictable pacing suitable for automated dubbing pipelines.
Fixed Voice Presets
SyncVoice offers eight specific voice presets (e.g., aria, nova, atlas). You cannot upload a custom voice or clone a specific talent’s voice for your project. Each voice supports all 23 languages, but the timbre remains fixed to the original licensed recording.
The choice is limited to the eight provided personas. There is no option for gradual tonal adjustments or creating a unique brand voice through cloning, unlike some consumer-grade services.
- Eight static voices: aria, nova, luna, sage, felix, milo, atlas, orion.
- No custom voice cloning or uploading of reference audio.
- All voices support the full range of 23 languages.
No Custom Emotion Tags
The API does not support SSML or explicit emotion tags (e.g., <emphasis>, <break>). You cannot programmatically force a voice to sound sad, excited, or angry. The emotional tone is inherent to the voice preset’s design, such as felix being energetic or luna being soft.
This limits fine-grained control over the delivery’s emotional nuance. For complex scripts requiring rapid emotional shifts, you may need to select different voices for different segments rather than adjusting a single voice’s tone on the fly.
- No SSML support for structural emphasis.
- Emotion is tied to the voice preset, not the input tags.
- No ability to dynamically adjust intensity or mood per sentence.
Future Trends
As neural networks evolve, TTS models are expected to deliver even more natural prosody and rhythm, reducing the robotic feel in long-form localization projects. Higher sampling rates and advanced format support like FLAC will enhance audio fidelity for professional dubbing workflows.
Improved Naturalness
Future iterations will likely refine intonation and pacing, making synthetic voices indistinguishable from human recordings in many contexts. This reduces the need for manual post-production editing in localization pipelines.
Higher Sampling Rates
Support for lossless formats such as FLAC ensures that audio remains pristine during distribution. Higher sampling rates will preserve subtle vocal nuances critical for documentary and e-learning content.
Better Language Support
While we currently support 23 languages, future updates aim to expand coverage to underserved regions. Improved handling of non-Latin scripts and dialectal variations will enhance global reach for localization agencies.
Security in TTS
Our TTS API is engineered for enterprise-grade data handling, ensuring that input text is used exclusively for audio synthesis and not repurposed for model training. While the generated audio is returned in the response without persistent storage, we retain request metadata—such as timestamps, character counts, and status codes—strictly for billing accuracy and audit trails.
This minimal data retention policy supports localization agencies processing sensitive scripts, as content is not archived after delivery. The absence of long-term text storage reduces exposure risks during high-volume dubbing workflows, providing a clear boundary between temporary processing and permanent storage.
Data Privacy
Input text is processed solely to generate the requested audio file and is not used for training purposes. Generated audio is returned in the response and not stored on our servers. Only request metadata—time, character count, and status—is retained for billing and audit purposes, ensuring minimal data footprint for high-volume localization projects.
This approach ensures that sensitive script content remains transient. Agencies can confidently submit proprietary content without fear of it being ingested into training datasets, maintaining strict control over intellectual property throughout the synthesis pipeline.
- Input text is used only for audio generation, not for training.
- Generated audio is not stored after the response is delivered.
- Request metadata is kept solely for billing and audit trails.
Watermarking
Every audio file generated by SyncVoice API contains an inaudible digital watermark that explicitly marks the content as synthetic speech. This watermark is embedded during synthesis and persists across all supported formats (mp3, wav, opus, flac, pcm), providing a consistent method for identifying AI-generated voiceovers in distributed media.
This feature is critical for localization agencies delivering content to platforms that require disclosure of synthetic media. The watermark operates independently of the voice model or language, ensuring reliable detection regardless of the audio characteristics or speed settings applied during generation.
- Inaudible watermark marks every file as synthetic speech.
- Watermark persists across all supported audio formats.
- Enables reliable detection of AI-generated voiceovers.
Synthetic Detection
The embedded watermark facilitates automated synthetic detection, allowing downstream systems to identify TTS-generated audio without relying on external machine learning models. This ensures compliance with emerging media transparency regulations, particularly for video dubbing where synthetic voices must be distinguished from human recordings.
Agencies can integrate detection tools to verify content origin before delivery to clients or broadcasters. The watermark’s resilience to common audio processing steps ensures that synthetic status remains identifiable even after compression or format conversion.
- Watermark enables reliable synthetic detection in downstream systems.
- Supports compliance with media transparency regulations.
- Resilient to common audio processing and compression.
Choosing the Right API
Selecting the correct API requires aligning technical constraints with project scale. SyncVoice API offers a straightforward consumption model based on character volume, ensuring predictable costs for high-volume localization pipelines without hidden fees or subscription locks.
Pricing Models
Our pricing is transparent: $0.01 per 1,000 characters, with failed requests refunded immediately. There are no monthly plans, minimum spends, or per-seat fees. Accounts start with a $0.50 trial valid for 7 days, and top-ups via USDT (TRC-20) are processed automatically within minutes.
- $0.01 per 1,000 characters
Voice Selection
We provide eight preset voices covering male and female ranges, suitable for diverse content types from news to storytelling. Each voice supports all 23 languages, allowing consistent tone across multilingual projects without needing additional voice talent.
- 8 presets: Aria, Nova, Luna, Sage, Felix, Milo, Atlas, Orion
Technical Support
Support is handled exclusively via email. We do not offer phone lines, live chat, or SLAs. For technical integration, documentation and cURL examples are provided to facilitate rapid implementation into your existing dubbing workflows.
Implementation Steps
Implementation begins by creating an account on playbd2.com via the /get-api-key/ page. You can sign in using Google or a standard email and password. This step establishes your unique identity within the SyncVoice API ecosystem, enabling access to your personal dashboard and billing information.
Account Creation
Register using Google or email/password on the /get-api-key/ page. Accounts are site-specific and require no initial commitment.
API Key Generation
Upon registration, your dashboard displays a single active API key starting with 'tts-'. This key authenticates all requests. You can generate a new key at any time, which immediately invalidates the previous one.
First Request
Send a POST request to https://api.playbd2.com/v1/audio/speech with your 'Authorization: Bearer' header. Include the text in the 'input' field and specify a voice model. The API returns the audio file directly in the response.
- Endpoint: POST https://api.playbd2.com/v1/audio/speech
- Auth: Authorization: Bearer tts-...
- Body: JSON with 'input', 'voice', 'model'
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. This scope targets core localization needs without unnecessary fragmentation.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- LibriTTS-R corpus (OpenSLR 141)
- TRON developer hub: TRC-20
- LibriVox public-domain audiobooks
- W3C WAI: making audio and video accessible
- Python documentation
- LibriTTS-R paper (arXiv 2305.18802)
- Creative Commons Attribution 4.0
- ISO 639 language codes
- Node.js documentation
- openai-python on GitHub
- FFmpeg documentation
