Frequently asked questions
What is voice cloning software?
Voice cloning software generates synthetic speech that mimics a specific individual’s unique vocal characteristics, often requiring a sample of their voice. It allows creators to replicate a real person’s tone and timbre for custom content, distinguishing it from standard text-to-speech systems that use fixed, generic preset voices.
Is this API voice cloning?
No. EchoVocal provides eight fixed preset voices (e.g., Aria, Nova, Felix) rather than cloning specific individuals. While generated audio contains an inaudible watermark marking it as synthetic, the system does not analyze or replicate a user’s unique vocal fingerprint or create a custom voice model.
How is it different from TTS?
This is a standard cloud Text-to-Speech API. You send text via a JSON POST request and receive an audio file. It does not clone specific voices or allow custom voice design. It uses eight licensed presets based on LibriTTS-R recordings, offering consistent, broadcast-ready synthetic speech without cloning variability.
Do you clone real voices?
No. We do not clone real voices, celebrities, or politicians. Our eight voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings. Each generated file carries an inaudible watermark identifying it as synthetic speech, ensuring transparency without replicating specific real-world individuals.
Can I clone my own voice?
No. We do not offer custom voice cloning, voice design, or the ability to upload your own voice samples. You must choose from our eight preset voices (e.g., Aria, Atlas, Luna). If you need a specific real-world voice, you must use a different service that supports cloning.
What voices are available?
We offer eight presets: Aria (warm narrator), Nova (bright), Luna (soft), Sage (calm), Felix (energetic), Milo (friendly), Atlas (neutral), and Orion (deep). These are based on licensed LibriTTS-R recordings. All voices can speak any of our 23 supported languages, with no custom cloning options.
How many 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, Thai, and others are not supported.
Is it expensive?
Pricing is $10 per 1 million characters ($0.01 per 1,000 characters). Failed requests are free. You can start with a $0.50 trial credit via Google sign-in. Top-ups are via USDT (TRC-20) with a $5 minimum. There are no monthly plans, subscriptions, or hidden fees.
Definition of Voice Cloning
Voice cloning in the synthetic audio industry refers to generating speech that mimics a specific person’s timbre, pitch, and cadence. It involves training a neural network on hours of reference audio to replicate a unique vocal identity rather than using a generic synthetic voice.
This API provides preset voices derived from licensed recordings, not custom cloning. While it generates high-fidelity synthetic speech, it does not offer the ability to upload a sample and create a personalized clone of a specific VTuber or streamer.
Synthetic voice generation
Synthetic voice generation uses deep learning models to convert text into audible speech. Unlike recording a human reading a script, the AI constructs phonemes and prosody dynamically. The resulting audio is entirely computer-generated, though it may carry an inaudible watermark to mark it as synthetic.
Text-to-speech vs. cloning
Standard Text-to-Speech (TTS) uses a fixed voice model trained on a broad dataset. Voice cloning creates a unique voice model specific to one individual. This API offers eight distinct preset voices (e.g., Aria, Felix) that are fixed. You cannot clone a new voice; you select from existing presets.
Use in VTubing
VTubers use TTS for off-screen commentary, automated responses, or multi-language localization of existing content. Since this API does not support real-time streaming, it is suited for pre-rendered video assets rather than live, interactive voice casting during a stream.
How It Works
EchoVocal’s TTS engine converts text into audio using neural network models trained on licensed recordings. The system processes your input text through these models to generate synthetic speech without requiring custom voice training or cloning infrastructure.
The synthesis pipeline handles character encoding, phoneme mapping, and waveform generation to produce broadcast-ready audio files. This process ensures consistent voice quality across different languages and preset characters, maintaining the distinct tonal characteristics of each voice model.
While not real-time streaming, the API delivers audio files efficiently. Typical generation speed matches real-time playback, with short phrases taking a few seconds. Under load, requests may queue, but the system processes them sequentially to maintain audio integrity.
Neural network models
EchoVocal uses neural network models trained on the LibriTTS-R dataset, derived from LibriVox public-domain audiobook readings. These models generate synthetic speech that retains natural prosody and intonation without requiring custom voice cloning.
The architecture supports eight distinct voice presets, each with unique tonal qualities. Models process text input to produce audio files, ensuring consistent quality across different languages and character sets.
- Trained on LibriTTS-R (CC BY 4.0) data
- Supports 23 languages with consistent quality
- Eight distinct voice presets available
Audio synthesis process
The synthesis process converts text input into audio files through a multi-step pipeline. Text is processed for character encoding, phoneme mapping, and waveform generation, resulting in high-quality synthetic speech.
Generated audio files are returned directly in the API response. The system does not store the audio files after delivery, ensuring that input text is used solely for synthesis and not for training purposes.
- Text input up to 4096 characters
- Supports MP3, WAV, Opus, FLAC, PCM formats
- Audio returned in API response
Real-time generation
Generation speed is approximately real-time, processing about 25 characters of text per second of audio. Short phrases typically take a few seconds to synthesize, while longer texts are split into parts and joined seamlessly.
Requests are processed sequentially, so under heavy load, a request may wait in a queue. This ensures audio integrity but means true real-time streaming is not available, distinguishing it from WebSocket-based solutions.
- ~25 characters per second generation speed
- Requests processed sequentially
- No streaming or WebSocket support
Voice Cloning Software vs. TTS
Many VTubers assume any synthetic voice is "cloning," but this API is standard Text-to-Speech. Cloning requires training on specific voice samples, whereas this service uses eight licensed presets derived from LibriTTS-R recordings. You select a preset like Aria or Felix rather than uploading your own voice model.
The distinction matters for ethics and workflow. Cloning often demands explicit consent from the voice source, while these presets are based on public-domain or licensed audiobook readings. This API generates speech from text using fixed neural models without mimicking a specific celebrity or creator's unique timbre.
Key differences
Voice cloning typically involves fine-tuning a model on a user's specific audio data to replicate their unique vocal characteristics. This API offers fixed presets like Nova or Orion, which are consistent across all users. There is no custom voice cloning feature; you cannot upload a sample to create a unique variant.
- No custom voice training or uploading of personal voice samples.
- Fixed presets rather than personalized voice models.
- No voice design or emotion tagging capabilities.
Similarities in output
Both technologies produce natural-sounding synthetic audio from text. The output quality is broadcast-ready, suitable for VTuber streams or indie game dialogue. Users cannot easily distinguish between cloned and preset voices without close listening, as both aim for high realism.
Cost implications
Cloning software often involves higher upfront costs for training models or premium subscription tiers. This API charges $0.01 per 1,000 characters with no minimum spend. Failed requests are free, and there are no monthly fees, making it cost-effective for high-volume text generation without the overhead of voice maintenance.
Licensing and Consent
The EchoVocal API delivers high-quality synthetic voices rooted in transparent licensing frameworks, ensuring clarity for VTubers and indie developers regarding usage rights. Our voice presets are derived from licensed recordings that provide a clear legal baseline for commercial and non-commercial content creation.
Understanding the distinction between the underlying license and the public domain source helps creators manage expectations and attribution correctly within their streaming workflows.
LibriTTS-R license
Our voice models are based on the LibriTTS-R dataset, which operates under a CC BY 4.0 license. This allows for broad usage, including commercial applications, provided appropriate attribution is maintained where required by your specific distribution channels.
The CC BY 4.0 framework ensures that the synthetic audio generated from these base recordings can be integrated into VTuber streams and video content without immediate royalty disputes.
- CC BY 4.0 licensed base recordings
- Allows commercial use with attribution
- Clear usage boundaries for synthetic output
LibriVox public domain
The source material originates from LibriVox, featuring public domain audiobook readings. This means the original spoken content is free of copyright restrictions, providing a stable foundation for our neural network training without encumbering the final synthetic audio with legacy copyright claims.
Public domain status simplifies rights management for creators who need predictable licensing for their VTuber avatars and stream clips.
- Source material from LibriVox
- Public domain audiobook readings
- Simplified rights management
Creator consent models
While the underlying recordings are public domain or licensed, the synthetic generation process introduces a new layer of consent. Each generated file carries an inaudible watermark marking it as synthetic speech, ensuring transparency with your audience.
This watermarking supports ethical voice usage by clearly distinguishing AI-generated audio from human performance, aiding in creator rights management and audience disclosure.
- Inaudible synthetic watermark in every file
- Supports audience transparency
- Clear distinction from human performance
Ethical Considerations
Voice synthesis raises distinct ethical questions regarding authenticity and attribution in the VTuber community. Unlike human performers, our synthetic voices lack biological origin but offer consistent, scalable performance without the fatigue constraints of live streaming sessions.
We prioritize transparency by embedding an inaudible watermark in every generated audio file. This technical marker identifies the output as synthetic speech, helping creators maintain trust with their audience by clearly distinguishing between human and AI-generated content.
Watermarking synthetic audio
Every audio file produced by our API contains a digital watermark. This invisible tag marks the audio as synthetic, providing a verifiable layer of authenticity without affecting sound quality or listener experience during playback.
This feature supports ethical disclosure by allowing platforms or creators to programmatically detect AI-generated content, ensuring that synthetic voices are not misrepresented as human recordings in commercial or broadcast contexts.
- Inaudible digital signature embedded in output audio
- Compatible with standard audio players
- Does not degrade voice clarity or emotional tone
Transparency in usage
Clear attribution helps maintain audience trust. When using preset voices like Aria or Felix, creators should disclose that the voice is AI-generated, especially if the voice is marketed as a specific character persona.
Transparency extends to the licensing source: our voices are derived from LibriTTS-R (CC BY 4.0) and LibriVox public-domain readings, ensuring that the underlying recordings respect creator rights and public domain standards.
Audience disclosure
Indie developers and streamers are encouraged to explicitly state when a voice is synthetic. This practice aligns with emerging community standards for AI-generated content in gaming and streaming platforms.
Disclosing synthetic voices helps audiences understand the nature of the performance, particularly when voices are used for emotional storytelling or character-driven narratives where human presence is expected.
Technical Requirements
Integration is straightforward for indie developers familiar with standard HTTP clients. The API mirrors the OpenAI speech endpoint shape, allowing you to reuse existing OpenAI SDKs by simply swapping the base URL to https://api.playabcd1234.com/v1. Authentication relies on a single Bearer token, ensuring secure, stateless communication without complex OAuth flows.
API integration
Send a POST request to /v1/audio/speech with your JSON payload. The system returns the audio file directly in the response body, eliminating the need for polling or secondary download endpoints. This synchronous pattern simplifies debugging and ensures immediate feedback on synthesis errors.
- Standard JSON request body
- Bearer token authentication
- Synchronous response delivery
Format compatibility
Select from mp3, wav, opus, flac, or pcm via the response_format field. MP3 offers efficient storage, while WAV and FLAC provide lossless quality for post-production editing. PCM is available for raw data processing pipelines.
- mp3, wav, opus, flac, pcm
- Lossless and lossy options
- Direct playback compatibility
Latency considerations
Synthesis is not real-time streaming; audio is generated after the request completes. Expect roughly one second of generation time per 25 characters of text. Under load, requests queue sequentially, so longer scripts may experience delays before the final audio file is returned.
- No WebSocket streaming
- Sequential processing under load
- ~25 chars/sec generation speed
Preset Voices Available
The API provides eight distinct preset voices, designed to cover a wide range of VTuber and streaming personas. These voices are generated from licensed recordings, ensuring broadcast-ready quality without the need for custom cloning infrastructure. Each voice is optimized for clarity and emotional range, making them suitable for character dialogue, narration, or live commentary.
All voices support 23 languages and can be adjusted for speed between 0.25x and 4.0x. This flexibility allows streamers to maintain consistent audio quality across different content types, from fast-paced gaming highlights to slower, story-driven segments. The voices are not custom-trained per user but are selected from a fixed, high-quality library.
Female voices: aria, nova, luna, sage
Aria offers a warm, clear narrator style, ideal for calm intros or educational streams. Nova is bright and upbeat, perfect for energetic gaming moments or quick updates. Luna provides a soft, storytelling tone, suitable for lore explanations or atmospheric content. Sage is calm and instructional, making it excellent for tutorials or guided streams.
- Aria: warm, clear narrator
- Nova: bright, upbeat
- Luna: soft, storytelling
- Sage: calm, instructional
Male voices: felix, milo, atlas, orion
Felix is energetic and promo-focused, great for hype moments or announcements. Milo is friendly and conversational, fitting well for casual chat or Q&A sessions. Atlas delivers a neutral, news-like tone, ideal for objective commentary or updates. Orion is deep and documentary-style, suited for serious narratives or dramatic reveals.
- Felix: energetic, promo
- Milo: friendly, conversational
- Atlas: neutral, news
- Orion: deep, documentary
Voice characteristics
Each voice is based on licensed recordings from LibriTTS-R (CC BY 4.0), sourced from LibriVox public-domain audiobooks. Every generated audio file includes an inaudible watermark to mark it as synthetic speech, aiding in transparency. Voices are not cloned from specific individuals but are presets available to all users. They can speak any of the 23 supported languages, with language detection or manual setting ensuring accurate pronunciation.

Language Support
EchoVocal supports 23 languages, ensuring your VTuber streams reach global audiences without needing separate voice assets. This coverage includes major markets like English, Japanese, Korean, and Spanish, alongside less common options like Swahili and Turkish, all processed through a single API endpoint.
The system handles language selection intelligently but allows for precise manual overrides when needed, giving streamers full control over vocal delivery in mixed-language segments.
23 languages supported
We support 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, allowing for versatile character voices without switching models.
- Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, Turkish
Auto-detection logic
If the language field is not set, the API guesses the language from the script. Latin script text defaults to English, simplifying requests for common English dialogue. Non-English Latin-script text (e.g., German or French) requires explicit language setting to avoid mispronunciation.
Manual language setting
Set the language parameter to an ISO code for precise control. This is essential for non-English Latin text or mixed-language streams where auto-detection might fail. Unsupported languages like Vietnamese, Thai, or Ukrainian are not processed and may produce inaccurate results.
- Use ISO codes for precise control
- Essential for non-English Latin text
- Avoids mispronunciation in mixed streams
Pricing Model
EchoVocal charges $0.01 per 1,000 characters, making it predictable for VTuber streamers managing long dialogue. Failed requests are free, ensuring you only pay for successful audio generation. Spaces and punctuation count toward the total, so plan your script lengths accordingly to avoid unexpected costs during live streams or video recordings.
Per character billing
The rate is $10 per 1 million characters. This linear model scales easily for indie developers who do not need monthly seats or minimum spend commitments. You can generate short sound effects or full-length audiobooks without changing billing structures.
- $0.01 per 1,000 characters
- Failed requests are free
- No monthly plans or subscriptions
Free trial availability
Sign in with Google to receive $0.50 of credit, valid for 7 days. This provides about 50,000 characters for testing voices like Aria or Orion before committing funds. Email-based accounts start without trial credit, so choose your sign-up method wisely.
- $0.50 credit for Google sign-ins
- Valid for 7 days
- Approx. 50,000 characters
USDT payment
Top-ups use USDT on the TRON network (TRC-20) only, with a minimum of $5. Credits appear automatically within minutes. There are no invoices, VAT charges, or credit card processing fees, keeping your overhead low for independent projects.
Balance does not expire, allowing you to accumulate funds during quieter production periods without losing value.
- TRC-20 USDT only
- Minimum $5 top-up
- Automatic credit within minutes
Data Privacy
EchoVocal’s TTS API treats your input text as ephemeral data. When you send a request, the text is processed solely to generate the corresponding audio file, which is then returned in the response. This text is not stored permanently in our database, ensuring that your scripts remain transient and secure during the synthesis process.
Input text usage
The input text is used exclusively for the immediate production of the audio asset. Once the audio file is generated and delivered to your client, the original text string is not retained for long-term storage or analysis. This ensures that your VTuber scripts or stream content remain private and are not archived indefinitely.
- Text is used only for immediate audio synthesis
- No long-term storage of input scripts
- Text is discarded after audio delivery
No training on data
Unlike many consumer AI services, EchoVocal does not use your input text or generated audio to train its neural network models. Your content remains your intellectual property, and the data you submit is not ingested into the model’s training dataset. This distinction is critical for indie developers who need to maintain exclusive rights to their voice assets and script libraries without fear of data leakage into public models.
- Input text is not used for model training
- Generated audio is not stored for training
- Content remains your exclusive property
Metadata retention
While the content of your requests is ephemeral, we retain specific metadata for billing and operational purposes. This includes the timestamp of the request, the character count processed, and the status of the synthesis (success or failure). This metadata is kept strictly to verify usage for your account and to ensure accurate billing, particularly for your USDT top-ups.
- Timestamps of requests
- Character counts processed
- Synthesis status (success/failure)
- Used for billing verification
Comparison with Other Solutions
We audit TTS APIs through the lens of VTuber workflows, focusing on voice cloning ethics and preset quality. While OpenAI and ElevenLabs dominate general markets, our API offers a distinct path for indie developers seeking broadcast-ready audio without the complexity of custom cloning.
This comparison highlights how our preset voices and transparent pricing model serve streamers who prioritize ethical licensing and predictable costs over real-time partial audio generation.
OpenAI API comparison
Our API mirrors the OpenAI speech endpoint structure, allowing you to swap the base URL and use the same SDKs. We support tts-1 and tts-1-hd models with identical request shapes. However, we do not offer streaming, SSML, or custom pronunciation lexicons, keeping the integration simple for VTuber overlays.
The OpenAI SDK works directly by pointing to https://api.playabcd1234.com/v1. This compatibility means your existing code requires minimal changes, focusing on voice selection rather than complex audio pipeline adjustments.
- Identical JSON request structure to OpenAI
- Supports tts-1 and tts-1-hd models
- No streaming or SSML support
ElevenLabs alternatives
Unlike ElevenLabs, we do not offer custom voice cloning or uploading your own voice. Our eight presets (Aria, Nova, Luna, etc.) are based on licensed LibriTTS-R recordings, ensuring clear consent models for VTuber creators. We provide fixed, predictable pricing without per-voice tier complexities.
We focus on preset quality and ethical licensing. While ElevenLabs offers cloning, our approach ensures every generated file carries an inaudible watermark, maintaining transparency about synthetic speech in your streams.
- No custom voice cloning
- Eight fixed presets
- Inaudible watermark in all audio
Cost analysis
Pricing is $10 per 1 million characters, or $0.01 per 1,000 characters. Failed requests are free, ensuring you only pay for successful synthesis. This model is straightforward compared to tiered subscription plans from competitors, making it ideal for indie developers with variable usage.
We accept USDT (TRC-20) for top-ups, with no monthly fees or seats. This crypto-only approach reduces friction for international streamers and keeps operational costs low, allowing competitive rates without hidden VAT or invoice delays.
- $0.01 per 1,000 characters
- Free failed requests
- USDT (TRC-20) payments only
Getting Started
Begin by visiting the /get-api-key/ page to create your EchoVocal account. You can sign up instantly using Google or create a standard email and password account. This step establishes your unique identity on playabcd1234.com and prepares your dashboard for usage tracking.
Sign up process
Google sign-ups receive an automatic $0.50 trial credit valid for seven days, sufficient for approximately 50,000 characters. Email-based accounts start with a zero balance, requiring you to top up before generating audio. There are no subscriptions or monthly plans to manage.
API key generation
Upon registration, your dashboard displays a single active API key starting with tts-. This key authenticates all requests to https://api.playabcd1234.com/v1. You can issue a new key at any time; the previous one stops immediately, ensuring secure access for your VTuber workflows.
First request
Use the official OpenAI SDK or a simple HTTP POST to https://api.playabcd1234.com/v1/audio/speech. Provide your model (tts-1 or tts-1-hd), selected voice, and input text under 4096 characters. The API returns the audio file directly, ready for your stream overlays.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. VTubers and streamers are turning our indie games into viral sensations, driving massive engagement through their unique and entertaining content.
23 languages
Any voice speaks any of these languages.
