Frequently asked questions
What is voice cloning?
Voice cloning creates a unique synthetic voice model trained on specific audio samples, allowing text-to-speech systems to mimic that exact speaker’s tone and timbre. It enables highly personalized audio generation that matches a particular person’s vocal characteristics rather than using a generic preset voice.
How is voice cloning different from TTS?
Standard TTS uses fixed voice presets like Aria or Felix, while voice cloning builds a custom model from your audio samples. Our API uses eight licensed presets and does not offer custom voice cloning, so you cannot upload your own voice to create a unique clone.
Is coqui tts voice cloning accurate?
Accuracy depends on the quality and length of the source audio used to train the clone. Since we do not offer voice cloning, this metric does not apply to our service. Our eight preset voices provide consistent, high-quality results without the variability of custom training.
Do I need consent to clone a voice?
Yes, typically you need consent to clone a recognizable person’s voice, especially for commercial use, as laws vary by region. Since we do not offer voice cloning, this concern is irrelevant for our preset voices, which are synthetic models based on licensed recordings.
Can I clone any voice?
You can clone any voice if you have sufficient audio data and the technical infrastructure to train the model. We do not offer voice cloning, so you cannot clone specific voices. You are limited to our eight preset voices, which are fixed and do not require custom training.
What is a synthetic watermark?
A synthetic watermark is an inaudible signal embedded in the generated audio file to mark it as synthetic speech. Every file we generate carries this watermark, helping identify that the audio was produced by our text-to-speech API rather than recorded by a human.
Are synthetic voices detectable?
Yes, synthetic voices can be detected using specialized AI tools or by listening for artifacts. Additionally, every audio file we generate contains an inaudible synthetic watermark that explicitly marks it as synthetic speech, providing a technical method for verification beyond human perception.
Is voice cloning expensive?
Voice cloning can be expensive due to training costs and storage requirements for custom models. We do not offer voice cloning, so you avoid these costs. Our pricing is simple: $10 per 1 million characters, with no fees for custom voice model training or maintenance.
Do you offer voice cloning?
No, we do not offer voice cloning. Our API uses eight fixed preset voices based on licensed recordings. You cannot upload audio samples to create a custom clone of a specific person. You select from Aria, Nova, Luna, Sage, Felix, Milo, Atlas, or Orion.
What voices do you offer instead?
We offer eight preset voices: Aria, Nova, Luna, Sage, Felix, Milo, Atlas, and Orion. These are based on licensed LibriVox recordings and cover various tones from warm narrators to deep documentaries. They support 23 languages and are ready to use immediately via our API.
Are the voices licensed?
Yes, our voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings from LibriVox. This ensures you can use the generated audio for commercial projects without worrying about copyright infringement, allowing you to focus on your animation story.
Can I use the voices commercially?
Yes, you can use the generated audio commercially. The voices are based on licensed or public-domain sources, and we do not claim ownership over the generated audio files. You pay only for the characters processed, with no monthly fees or restrictions on commercial usage.
What is the difference between presets and cloning?
Presets are fixed voice models like Aria or Felix that you select from a list. Cloning creates a custom model trained on your specific audio samples. We offer only presets, so you cannot create a unique clone of a specific person, but you get consistent, high-quality results.
Do preset voices sound robotic?
Our preset voices are designed to sound natural and expressive, not robotic. They are based on licensed recordings and engineered for clarity and emotion. While they lack the specific nuance of a cloned voice, they provide high-quality speech suitable for animation and cartoons without sounding artificial.
Defining Voice Cloning
Voice cloning creates a digital model of a specific voice using sample audio, allowing text-to-speech engines to mimic that unique timbre. In animation, this promises consistent character voices across episodes without requiring the same actor for every line.
Unlike preset voices that offer a fixed set of tones, cloning aims to replicate the nuances of a human speaker. This section defines the core concepts and data needs before comparing them to our preset library.
What is Voice Cloning?
Voice cloning is a subset of TTS that generates synthetic speech matching a specific person’s vocal characteristics. It requires training a model on existing audio samples to learn pitch, rhythm, and tone.
How Cloning Works Technically
Technically, cloning involves feeding audio data into a neural network that learns the speaker’s acoustic features. The model then applies these learned patterns to new text inputs, generating audio that sounds like the original speaker.
Data Requirements for Cloning
High-quality cloning typically demands hours of diverse speech data from the target voice. The more varied the samples, the more natural and consistent the cloned voice becomes in different contexts.
Coqui TTS and Cloning
Coqui TTS was a prominent open-source text-to-speech project known for its high-quality preset voices and modular architecture. It gained popularity among developers for its ability to generate natural-sounding speech from text inputs using neural networks trained on large datasets of licensed recordings.
The project offered various voice models, including the widely used VITS architecture, which allowed for expressive and realistic audio generation. Its open-source nature made it a go-to choice for startups experimenting with voice synthesis without licensing fees.
What is Coqui TTS?
Coqui TTS was an open-source text-to-speech engine developed by Coqui AI. It provided high-quality preset voices trained on licensed datasets, enabling developers to generate realistic speech from text inputs. Its modular design allowed for easy integration into various applications, from animation to voice assistants.
The system relied on neural networks, particularly the VITS architecture, to produce natural-sounding audio. It was popular for its balance of quality and accessibility, making it a staple for developers experimenting with voice synthesis without significant licensing overhead.
- Open-source architecture
- VITS-based voice models
- High-quality preset voices
- Easy integration for developers
Cloning Capabilities in Coqui
Coqui TTS supported voice cloning through its model architecture, allowing users to generate speech in the style of a specific voice. This was achieved by fine-tuning the model on a small dataset of the target voice, enabling the creation of custom voice clones for various applications.
While Coqui offered cloning capabilities, it required additional processing and data preparation. Users needed to provide a sample of the target voice, which the model would then use to generate speech in that voice's style. This process was more complex than using preset voices.
- Voice cloning via fine-tuning
- Requires target voice samples
- More complex setup
- Custom voice generation
Comparison to Preset Models
Preset models in Coqui TTS were pre-trained and ready to use, offering consistent quality across different applications. They were ideal for projects where a specific, consistent voice was needed without additional processing. Presets were easier to integrate and required less data preparation.
In contrast, voice cloning allowed for custom voices but demanded more effort in terms of data collection and model fine-tuning. While cloning provided unique voice options, presets were more efficient for projects requiring consistent, high-quality speech without the need for customization.
For animation startups, presets offered a reliable and straightforward solution for generating character voices. Cloning was better suited for projects requiring unique, personalized voices, but at the cost of additional time and resources for setup and processing.
- Preset models: consistent quality
- Cloning: requires fine-tuning
- Presets: easier integration
- Cloning: more complex setup
Cloning vs. Preset Voices
To create a unique character voice tailored to a specific show identity, custom cloning captures nuanced speech patterns that generic models might miss. This approach allows for highly personalized audio that aligns precisely with a character's personality, ensuring consistency across episodes without relying on standard presets.
Pros of Custom Cloning
Cloning allows for unique character voices tailored to specific show identities. It captures nuanced speech patterns that generic models might miss. (Note: This is a general description of cloning, not a feature of our API.)
Pros of Preset Voices
Our preset library offers eight distinct voices, each with a specific tone like warm narration or energetic promo. These voices are built from licensed LibriVox recordings, providing immediate consistency and copyright safety without the need for additional voice design or training data.
Cost and Complexity Differences
Custom cloning typically requires gathering high-quality audio samples and running training processes, adding time and technical overhead. In contrast, our preset voices are ready instantly via API, eliminating training costs. You pay only for characters generated, with no minimum spend or subscription fees.
Consent and Licensing
Voice cloning often raises legal questions about who owns a character's voice. Our approach sidesteps this by using preset voices built from licensed recordings, ensuring you avoid complex consent negotiations for every character.
We use recordings from LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobooks. This means the underlying voice data is already cleared for commercial use, reducing your legal overhead significantly.
Why Consent Matters
In animation, using a celebrity or unique voice without permission can lead to costly lawsuits. By using generic, pre-cleared voices, you eliminate the need to track down individual consent forms for every character you create.
- No celebrity voice rights issues
- Reduced legal overhead
Licensing Licensed Recordings
Our voices are based on LibriTTS-R (CC BY 4.0) and LibriVox public-domain audiobooks. These sources provide a robust legal foundation, allowing you to generate synthetic speech without worrying about underlying copyright claims on the voice source itself.
- LibriTTS-R: CC BY 4.0
- LibriVox: Public Domain
Commercial Use Rights
Since the source recordings are cleared for commercial use, you can use the generated audio in your cartoons, games, or ads without additional licensing fees. This makes Voxyto ideal for startups needing scalable, legally safe voice assets.
Note: We do not offer custom voice cloning, so you cannot replicate a specific person's voice without their direct consent. Our preset library covers 8 distinct styles, all commercially safe.
- Safe for commercial cartoons
- No extra licensing fees
Synthetic Speech Detection
Every audio file generated by Voxyto contains an inaudible digital watermark that marks it as synthetic speech. This subtle signal is embedded directly into the audio waveform, making it detectable by specialized tools but imperceptible to human listeners during normal playback.
The watermark serves as a clear indicator that the voice is AI-generated, providing transparency about the origin of the audio. It helps distinguish our licensed LibriVox-based voices from human recordings without affecting the listening experience for your audience.
What is a Synthetic Watermark?
A synthetic watermark is a unique, inaudible digital signature embedded within the audio file during generation. It is not a visible logo or a spoken cue, but a data pattern that audio detection algorithms can identify to confirm the file’s AI origin.
Why Watermarks Are Used?
Watermarks provide transparency in media production, helping platforms and distributors identify synthetic content. This is increasingly important for content moderation, copyright tracking, and maintaining audience trust by clearly distinguishing AI voices from human performances.
- Enables automated detection by media platforms
- Maintains transparency for end-users
- Supports copyright and attribution tracking
Implications for Distribution
For animation startups, this means your character voices are clearly identifiable as synthetic. While this aids in content classification, it may require disclosure on platforms that mandate labeling of AI-generated media. The watermark does not affect audio quality or compatibility with standard video editing workflows.
Use Cases in Animation
Animation startups need consistent character voices. Cloning offers unique sounds for main characters, but we use preset voices built from licensed recordings. This ensures copyright safety without the complexity of training custom models on your data. Our API lets you generate audio directly from text using eight distinct presets, covering various tones from energetic to calm.
Preset voices are ideal for pilots or quick tests. You can switch voices easily between episodes without retraining. This approach saves time and keeps production simple, allowing your team to focus on storytelling rather than managing complex voice models. Our voices are based on licensed LibriVox recordings, ensuring you can produce content without legal headaches.
When Cloning Makes Sense
Cloning shines when you need a unique, proprietary sound that no preset can replicate. It is ideal for main characters where vocal identity drives the brand, allowing you to train a model on a specific actor’s voice for long-form content. (Note: Cloning is not offered by our API.)
- Best for long-term series with consistent vocal needs
- Requires significant data and training time
- Not available in our current preset-based API
When Presets Are Better
Preset voices offer immediate consistency and speed. You can generate audio for a pilot episode quickly to test market response. This approach is cost-effective for startups that need to iterate rapidly without the overhead of voice model management.
Our eight presets cover a wide range of tones. From warm narrators to energetic promo voices, you can find a match for most characters. This flexibility allows you to experiment with different vocal styles without technical debt.
- Immediate availability with no training required
- Lower cost per request
- Easy to switch between characters
Examples from Cartoon Production
Use presets for a quick pilot episode to test market response. Switch to cloning only if the character becomes a series staple, justifying the initial setup time against the long-term need for consistent, copyright-safe audio generation. (Note: Cloning is not offered by our API.)
For a children's show, use a warm preset for the main character and a bright preset for the sidekick. This creates distinct identities without needing custom voice models. Our API handles the generation, so you can focus on the script and animation.
- Test characters with different presets
- Maintain consistency across episodes
- Reduce legal risks with licensed voices

Technical Limitations
The API uses eight fixed voice presets rather than generating unique models. You cannot upload a sample recording to create a new voice model, so your characters must fit within the existing library. This eliminates the need for complex training pipelines but limits vocal variety to the provided options.
Each preset is a distinct character type, such as a warm narrator or deep documentary voice. While this ensures consistency, it means you cannot create a one-off voice for a specific background character or niche role. The solution is to mix and match the eight available voices to build your cast.
No Custom Voice Cloning Here
We do not offer custom voice cloning. You cannot upload a sample recording to create a new voice model. This means your characters must fit within the eight available presets.
Fixed Voice Library
The voice library is static and does not expand with new uploads. You are limited to the eight presets: Aria, Nova, Luna, Sage, Felix, Milo, Atlas, and Orion. These voices are based on licensed recordings and remain consistent across all generations.
No Voice Design Tools
There are no tools to design or tweak voice characteristics beyond speed and language. You cannot adjust pitch, timbre, or emotion tags. Each voice has a fixed personality, such as 'bright and upbeat' or 'calm and instructional', which you select by name in the API request.
This simplicity reduces integration complexity but means you cannot fine-tune a voice to match a specific animation style or character arc precisely.
- No pitch or timbre adjustments
- No emotion tags
- No custom voice design interface
Audio Quality and Realism
Our eight preset voices deliver clear, consistent audio ideal for animation workflows. Each voice, derived from licensed LibriVox recordings, offers a distinct character type—from the warm narrator to the deep documentary tone—ensuring your cartoon characters sound professional without legal headaches.
The synthetic nature of the audio is marked by an inaudible watermark, confirming its origin as generated speech. This transparency helps maintain trust in your production's assets while keeping the focus on storytelling rather than technical perfection.
Voice Characteristics
Voices like Aria or Felix provide specific tonal qualities that fit cartoon archetypes. They are designed for clarity and consistency, avoiding the unpredictable variations often found in untrained cloning models.
Each preset speaks any of our 23 supported languages naturally. The base recordings come from public-domain audiobooks, ensuring a clean, noise-free audio profile suitable for direct integration into animation pipelines.
- Female presets: Aria, Nova, Luna, Sage
- Male presets: Felix, Milo, Atlas, Orion
- Clear, consistent timbre across all languages
Naturalness of Speech
Speech flows naturally with standard intonation patterns. The engine handles text up to 4096 characters per request, maintaining coherence without abrupt cuts or robotic pauses typical of older TTS systems.
Generation speed is roughly real-time, allowing for quick iterative testing. This efficiency lets animators adjust scripts and listen to results in near real-time, speeding up the production cycle.
- Real-time generation speed
- Consistent intonation
- No robotic pauses
Emotion and Tone Limits
Voices do not support emotion tags or dynamic tone shifts. Each preset maintains a steady emotional baseline, such as 'upbeat' for Nova or 'calm' for Sage, regardless of the text content.
For nuanced emotional performance, you must rely on script context and voice selection rather than technical parameters. This simplicity ensures predictable audio output for consistent character voices.
- No emotion tags
- Steady emotional baseline
- Script-driven tone
Choosing the Right Approach
Choosing the right voice strategy depends on your animation studio's specific constraints. If you need consistent, copyright-safe character voices without legal overhead, preset voices offer a reliable baseline. If you require unique timbres for distinct characters and can manage the licensing complexity, cloning might be worth the effort.
For startups balancing budget and speed, presets provide immediate scalability. Cloning introduces variable costs and consent management, which can slow production pipelines. Evaluate whether the unique character of a cloned voice justifies the additional administrative and technical overhead for your specific series.
Assessing Your Project Needs
Start by listing your character count and dialogue volume. If you have many minor characters with short lines, preset voices like Aria or Felix are efficient. For lead roles requiring distinct emotional range, consider if a cloned voice adds necessary narrative value or if a preset like Luna suffices.
Evaluate your distribution channels. If you are streaming globally, synthetic watermarks ensure clear identification. Preset voices built on licensed recordings simplify rights management, allowing you to focus on story rather than clearance documentation for each unique voice source.
- High character count favors presets for consistency.
- Limited legal team prefers licensed preset voices.
- Distinct lead characters may justify cloning effort.
Budget Considerations
Preset voices operate on a simple per-character model at $0.01 per 1,000 characters. This predictable cost structure helps startups forecast expenses accurately without hidden fees. Failed requests are free, reducing risk during development phases.
Cloning requires upfront investment in voice talent and legal consent. While our API does not offer custom cloning, external cloning services add variable costs. For tight budgets, leveraging our 8 preset voices maximizes creative output per dollar spent.
- Predictable pricing: $0.01 per 1,000 characters.
- No minimum spend or monthly plans.
- Failed requests are free.
Timeline and Workflow
Presets enable immediate integration. With our API, you can generate audio in seconds, fitting into agile production cycles. The 60 requests per minute limit supports steady workflow without complex queuing for most small-to-medium projects.
Cloning workflows often involve recording sessions, rights clearance, and model training, adding weeks to pre-production. If your timeline is tight, presets allow you to prototype episodes quickly and iterate based on feedback without waiting for voice asset preparation.
- Immediate generation: seconds per request.
- No waiting for model training.
- Fits agile production cycles.
Future of Voice Cloning
The TTS landscape is shifting from static presets toward more adaptive, context-aware synthesis. While true voice cloning remains a distinct technology, current trends focus on enhancing the expressiveness of existing models rather than requiring unique training data for every character. This lowers the barrier for animation startups needing consistent audio without managing complex cloning pipelines.
Trends in TTS Technology
Generative models are becoming more efficient, allowing for faster inference and lower latency. We expect broader support for multi-speaker capabilities within single models, reducing the need for separate voice files for minor characters. (This is an industry trend, not a current feature.)
Contextual awareness is improving, enabling voices to adapt slightly to narrative tone without full cloning. This bridges the gap between rigid presets and expensive custom cloning.
- Increased model efficiency
- Better contextual adaptability
Evolution of Synthetic Voices
Synthetic voices are moving beyond simple text-to-audio conversion to include nuanced emotional delivery. While we currently offer preset emotions via voice selection, future iterations may allow finer control over tone and pacing within the same voice identity.
The distinction between 'preset' and 'cloned' voices may blur as models become more versatile, allowing high-fidelity results from minimal input data.
What to Expect Next
We anticipate more standardized interfaces for voice management, making it easier to swap characters or update voices without retraining. Expect improved language support and better handling of long-form content without quality degradation. (This is an industry trend, not a current feature.)
While we do not offer custom cloning now, the industry trend suggests that lightweight cloning will become more accessible. For now, our preset library offers a reliable, copyright-safe foundation for your animation needs.
Focus on narrative consistency using our eight established voices, which are built from licensed recordings and ready for immediate use in production.
- Standardized voice management
- Accessible lightweight cloning
Summary and Recommendations
For animation startups, consistency and legal safety often outweigh the novelty of custom cloning. Our preset library provides eight distinct, copyright-safe voices built from licensed LibriVox recordings, eliminating the need for complex consent negotiations or large training datasets.
Key Takeaways
Preset voices offer immediate deployment with zero training overhead. Each voice is clear, consistent, and legally secure for commercial use. You avoid the hidden costs of data curation and quality control associated with custom cloning.
- Eight preset voices cover most character archetypes.
- No consent paperwork required for base voices.
- Instant API integration via standard JSON requests.
Best Practices for Startups
Start with our preset voices to validate your story’s audio direction. Use the API’s speed and format options to prototype quickly. Only consider custom cloning if a specific, unique voice becomes a core brand asset after initial success.
- Test voices with short scripts first.
- Monitor character usage to ensure consistency.
- Keep workflows simple with direct API calls.
Final Thoughts on Cloning
Voice cloning is powerful but often unnecessary for early-stage animation. Our solution prioritizes reliability, legal clarity, and ease of use. Focus your resources on storytelling rather than managing voice datasets and licensing risks.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. You receive the full voiceover asset to download and integrate into your animation software.
23 languages
Any voice speaks any of these languages.
