Frequently asked questions
Is the voice generation truly realistic for gaming streams?
Each voice carries an inaudible watermark marking it as synthetic speech. The eight presets range from warm narrators to deep documentary styles, offering clear audio suitable for voiceover work. They are distinct synthetic voices rather than indistinguishable human recordings, ensuring transparency for your audience.
Can I use these voices for commercial content?
Yes. The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings. There are no monthly plans, seats, or minimum spend requirements. You can generate audio for commercial streams without additional licensing fees beyond the per-character usage cost.
How many languages does the API support?
The API supports 23 languages, including English, Spanish, German, Japanese, Korean, and Arabic. Any voice can speak any of these languages. If the language is not specified, it is guessed from the script, though setting it explicitly is recommended for non-English Latin text.
What happens if the synthesis fails?
Failed requests are free. If synthesis fails, the charge is returned in full. You are only billed for successful audio generations. This ensures you do not pay for errors or incomplete outputs, protecting your budget during high-volume streaming preparation workflows.
Do I need to install any software to use the API?
No software installation is required. It is a cloud API accessible via standard HTTP requests. You can use the official OpenAI SDKs by changing the base URL or send direct POST requests. No mobile SDK or WordPress plugin is needed to integrate it into your workflow.
Are the voices based on real human recordings?
Yes. Every generated file carries an inaudible watermark that marks it as synthetic speech. The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0) and public-domain audiobook readings from LibriVox, providing a foundation of real human speech patterns for each preset.
How long does it take to generate audio?
Typical speed is roughly real time, generating about 25 characters per second. Short phrases take a few seconds, while longer texts process sequentially. Under load, requests may wait in a queue, but the output generally matches the duration of the spoken text.
Can I change my API key if it gets compromised?
Yes. You can issue a new key at any time from your account dashboard. When a new key is issued, the old one stops working immediately. This allows for instant security response without downtime, ensuring your streaming assets remain protected if credentials are exposed.
Is there a limit to how long my voice clips can be?
Each request accepts up to 4096 characters of input text. For longer scripts, you must split the text into parts and join the resulting audio files. This limit applies per API call, so complex narratives require multiple requests to generate the complete audio clip.
Do I need to credit the voice artist?
The source material uses LibriTTS-R (CC BY 4.0) and public-domain readings. While attribution is good practice for CC BY 4.0 content, the API itself does not mandate specific artist credits in your generated audio files. The inaudible watermark identifies the speech as synthetic rather than human-recorded.
What audio formats are available for download?
You can choose between mp3, wav, opus, flac, and pcm formats via the response_format parameter. These options allow you to select the best balance of quality and file size for your streaming platform, whether you need lossless FLAC or compressed MP3 for live use.
Can I adjust the speed of the voice?
Yes. The speed parameter accepts values between 0.25 and 4.0. This allows you to slow down narration for clarity or speed up promotional content. Adjusting speed does not affect the character count billing, as charges are based on the input text length.
Is there a free trial available?
Yes. Signing in with Google provides $0.50 of credit, valid for 7 days, covering approximately 50,000 characters. Accounts created with email and password start without trial credit. This allows you to test voice quality and API integration before committing to USDT top-ups.
How do I know which voice fits my brand?
We provide detailed reviews of the eight presets. Aria offers a warm narrator tone, while Felix is energetic for promos. Atlas provides a neutral news style, and Orion delivers a deep documentary voice. Analyze these specific traits to match your stream's emotional range and clarity needs.
First request
curl
curl https://api.marsgaming.info/v1/audio/speech \
-H "Authorization: Bearer $TTS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"tts-1","voice":"aria","input":"Hello from the API.","response_format":"mp3"}' \
--output hello.mp3Why Synthetic Voices Work for Streaming
Synthetic voices provide consistent audio levels and tone across hours of streaming, eliminating the natural variations that occur with human speech. This stability ensures your audience hears a uniform voice, regardless of session length or time of day.
Unlike human vocalists, synthetic voices do not suffer from vocal fatigue. You can generate content for long streams without the voice degrading in quality or energy, ensuring every line is delivered with the same clarity and intent.
Consistency for Long Sessions
Streaming often involves marathon sessions where vocal consistency is critical. Synthetic voices maintain the same pitch, pace, and emotional tone from the first minute to the last, preventing the listener from experiencing jarring shifts in speaker energy or volume that can break immersion.
This consistency is particularly valuable for background narration or continuous commentary, where a steady, predictable audio profile helps maintain viewer engagement without demanding constant attention to vocal nuances.
- Uniform tone across 4+ hour streams
- No volume or pitch drift over time
- Predictable audio profiles for background narration
No Vocal Fatigue
Human voice quality degrades after extended use due to muscle fatigue. Synthetic voices generated via API do not experience this physiological limitation. Each request is processed independently, delivering the same high-quality audio regardless of whether it is the first or the thousandth request of the day.
This means you can schedule voiceovers for late-night streams or early-morning posts without worrying about a 'tired' voice. The output remains crisp and clear, maintaining the intended emotional impact of the script.
- No degradation after hours of use
- Consistent quality for scheduled posts
- No need for vocal rest periods
Instant Content Production
With an API-based workflow, you can generate audio files on demand. This allows for rapid iteration of content, enabling you to update stream overlays or create quick reaction clips without waiting for a recording session to finish.
The ability to generate audio instantly means you can respond to live events with voiceovers in real-time, provided the character limit and rate limits are respected. This agility is crucial for dynamic content creation where timing is key.
- Generate audio on demand via API
- Rapid iteration of content
- Quick response to live events
The 8 Voice Presets Compared
Choosing the right synthetic voice is critical for maintaining viewer engagement during long stream sessions. Each of the eight presets offers distinct vocal characteristics, from warm narration to energetic promo styles, allowing creators to match their avatar’s persona precisely. Understanding these differences ensures your audio identity remains consistent and recognizable across diverse content types.
The following breakdown compares the four female and four male voices based on clarity, emotional range, and technical suitability for gaming commentary, storytelling, and news updates. This analysis helps you select the voice for your specific streaming niche.
Female: Aria, Nova, Luna, Sage
Aria provides a warm, clear narrator tone ideal for structured commentary. Nova offers a bright, upbeat delivery perfect for high-energy gaming moments. Luna’s soft, storytelling quality suits narrative-driven streams, while Sage delivers calm, instructional clarity for tutorials and guides.
- Aria: Warm, clear narrator
- Nova: Bright, upbeat energy
- Luna: Soft, storytelling tone
- Sage: Calm, instructional clarity
Male: Felix, Milo, Atlas, Orion
Felix brings energetic, promo-style delivery suitable for highlights and promotions. Milo offers a friendly, conversational tone for interactive streams. Atlas provides a neutral, news-style delivery for updates, while Orion’s deep, documentary voice adds gravity to serious content.
- Felix: Energetic, promo style
- Milo: Friendly, conversational
- Atlas: Neutral, news style
- Orion: Deep, documentary tone
Voice Personality Breakdown
Selecting a voice depends on matching vocal traits to your avatar’s persona. Use Aria or Atlas for consistent, clear communication. Choose Nova or Felix for dynamic, high-energy segments. Opt for Luna or Orion when emotional depth or gravity is required. Sage and Milo work best for casual, instructional content.
Technical Performance Metrics
Synthesis speed averages roughly real-time, generating about 25 characters of text per second of audio. Short phrases may take a few seconds to process, while longer texts scale linearly. Under heavy load, requests queue sequentially, so expect slight delays during peak usage.
Each API call accepts up to 4096 characters of input. For longer scripts, split the text into chunks, generate each segment, and join the resulting audio files. Failed requests incur no charge, ensuring you only pay for successfully generated audio.
Generation Speed vs. Real-Time
Processing is sequential, meaning your request waits in line if the server is busy. While typical generation matches real-time playback, short clips often require a few seconds of overhead. Plan your stream overlays accordingly, as instant playback is not guaranteed during high traffic.
Character Limit Handling
The API enforces a strict 4096-character limit per request. To produce longer content, divide your script into multiple requests. The client-side logic should concatenate the returned audio files seamlessly. This approach ensures consistent quality without hitting endpoint limits.
- Maximum input: 4096 characters per request
- Characters include spaces and punctuation
- Concatenate outputs for longer scripts
Audio Format Options (MP3, WAV, OPUS)
Choose between MP3, WAV, OPUS, FLAC, or PCM formats. MP3 offers efficient compression, while WAV provides lossless quality for editing. OPUS balances size and quality. Select the format that best fits your streaming software's requirements.
- MP3: Efficient, widely supported
- WAV: Lossless, high quality
- OPUS: Low bandwidth friendly
Language Support for Global Audiences
Synthetic voice generation supports 23 languages, allowing streamers to reach global audiences without hiring native voice actors. The API handles Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swahili, Swedish, and Turkish.
Every voice preset speaks all 23 languages with equal fidelity, ensuring consistent brand identity across international streams. There are no language-specific voice variants; the same vocal timbre applies regardless of the spoken tongue.
23 Supported Languages
The service supports 23 specific 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. Languages not on this list, such as Vietnamese, Indonesian, Thai, Ukrainian, Romanian, Czech, and Bengali, are not supported.
All eight voice presets function identically across all supported languages. A streamer can switch from English commentary to Spanish or Japanese announcements without changing voice settings, maintaining audio consistency.
- 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 vs. Manual Setting
The API automatically detects language from the input script. For Latin script text, it defaults to English. If you are generating Latin-script text in another language (e.g., Dutch or Portuguese), you must explicitly set the language field in the JSON body.
Failure to specify the language for non-English Latin text may result in incorrect pronunciation or accent. For non-Latin scripts like Arabic or Japanese, auto-detection usually works accurately.
Latin Script Default Behavior
When the language parameter is omitted, the system assumes English for any text using Latin characters. This simplifies workflows for English-dominant streamers but requires manual configuration for multilingual content using Latin scripts.
Example: Sending French text without language: "fr" will likely result in an English pronunciation of the French words.
Audio Quality and Watermarking
Every generated audio file carries an inaudible watermark that marks it as synthetic speech. This digital signature allows platforms to identify AI-generated content without affecting the listening experience, ensuring transparency with your audience.
The voices are based on licensed recordings from LibriTTS-R (CC BY 4.0), derived from LibriVox public-domain audiobook readings. This foundation provides a clean, high-fidelity audio source.
Inaudible Synthetic Watermark
The watermark is embedded directly into the audio waveform. It is designed to be inaudible to human listeners, preserving the quality of your stream while providing a clear marker of synthetic origin.
This feature is crucial for platforms that require disclosure of AI voices. It helps maintain trust with your viewers by allowing them to identify the synthetic nature of the narration if they choose to look for it.
- Compatible with MP3, WAV, OPUS, FLAC, and PCM formats.
- Does not degrade audio fidelity or add noticeable noise.
- Standard across all voice presets and languages.
Source Material: LibriTTS-R
Our engine uses recordings from the LibriTTS-R dataset, which consists of public-domain audiobook readings. This ensures a high baseline of audio quality and natural prosody.
Because the source material is public domain, we can offer a transparent and legally sound foundation for your content creation.
- Derived from LibriVox public-domain readings.
- Clean audio source for consistent quality.
- No hidden training data restrictions.
License Implications for Streamers
The CC BY 4.0 license applies to the source recordings. While this generally allows commercial use, you should verify the specific terms with your platform’s policies.
The synthetic nature of the output means you are creating a derivative work. The watermark helps distinguish this from human-recorded audio, which may be important for royalty claims or content ID systems.
We do not provide legal advice. Please consult the official license text for LibriTTS-R to ensure compliance with your specific use case.
- Verify platform-specific AI disclosure rules.
- Check CC BY 4.0 commercial use terms.
- Watermark aids in content identification.

Pricing Structure for Creators
Pricing is straightforward: $0.01 per 1,000 characters, including spaces. Failed requests incur no charge, ensuring you only pay for successful audio generation.
Cost Per Million Characters
Each million characters costs $10. This flat rate applies regardless of voice or format, making budgeting predictable for high-volume streamers.
Trial Credit Details
Signing in with Google provides $0.50 credit (approx. 50,000 characters) valid for 7 days. Email/password accounts start with zero balance.
USDT Top-Up Process
Add funds via USDT on the TRON network (TRC-20). Minimum top-up is $5, with automatic crediting within minutes. No cards or subscriptions.
- TRC-20 network only
- Minimum $5 top-up
- Automatic credit within minutes
- No monthly plans or invoices
API Integration for Stream Overlays
VoxHype’s API is designed for seamless integration with popular streaming tools. By adjusting the base_url in the official OpenAI SDKs to https://api.marsgaming.info/v1, you can use existing client libraries to handle authentication and audio retrieval without learning a new protocol.
OpenAI SDK Compatibility
The endpoint mirrors the OpenAI speech structure, allowing you to reuse code from other TTS services. Simply swap the base URL and inject your tts-... API key. This compatibility ensures rapid deployment for overlays or chatbot responses without custom wrapper code.
Rate Limits Explained
Each API key is limited to 60 requests per minute. New accounts start with a stricter limit of 10 requests per minute until the first top-up. Under heavy load, requests queue sequentially, so plan your burst traffic accordingly.
Simple HTTP Requests
For lightweight integrations, send a POST request to /v1/audio/speech with JSON body containing model, input, voice, and response_format. The API returns the audio file directly, ready to be piped into your stream software’s audio input source.
- POST https://api.marsgaming.info/v1/audio/speech
- Headers: Authorization: Bearer tts-...
- Body: {"model": "tts-1", "voice": "nova", "input": "Hello viewers", "response_format": "mp3"}
Common Pitfalls in Voice Selection
Selecting a voice that clashes with your avatar’s visual persona breaks immersion. A deep, documentary-style voice like Orion feels mismatched for a bright, energetic mascot, while the warm narrator Aria may lack the punch needed for high-stakes competitive commentary. Align vocal tone with character design to maintain credibility.
Monotone delivery often stems from static speed settings. Even expressive voices sound flat if read at a constant 1.0x pace. Varying intonation manually or adjusting speed slightly helps, but rely on the voice’s inherent range rather than forcing artificial emphasis through text formatting.
Matching Voice to Avatar Persona
Ensure the voice matches your on-screen character. Aria suits calm guides, while Felix’s energetic tone fits promo-style streams. Mismatched pairings distract viewers who expect visual-audio consistency.
Use Atlas for neutral news updates or Orion for deep lore explanations. Avoid using Luna for high-energy gameplay moments where her soft storytelling tone might dampen excitement.
- Match voice warmth to avatar personality
- Avoid deep voices for bright mascots
- Use energetic voices for hype moments
Avoiding Monotone Delivery
Static speed creates robotic rhythm. Even with expressive presets, reading at exactly 1.0x can sound dull. Slight variations in pacing keep the audio engaging.
The API allows speed adjustments from 0.25x to 4.0x. Use these to emphasize key moments, but avoid extreme settings that distort clarity.
- Vary speed slightly for natural flow
- Avoid 1.0x as the only setting
- Use speed to highlight key info
Speed Adjustments for Pacing
Fast-paced streams benefit from higher speeds (1.2x–1.5x), while storytelling segments work better at 0.8x–1.0x. Test different speeds with your specific content type.
Remember that requests are processed sequentially. During high traffic, longer texts may queue, affecting real-time responsiveness. Plan buffer times for live overlays.
- Use 1.2x–1.5x for fast content
- Use 0.8x–1.0x for storytelling
- Account for queue delays in live streams
Final Verdict for Content Creators
For streamers seeking dynamic energy, Felix delivers the punch needed for high-stakes gaming commentary, while Luna’s soft tone suits immersive storytelling. Atlas provides the neutral clarity required for news updates, and Orion’s depth adds gravity to documentary-style segments.
These selections are based on voice personality breakdowns and clarity metrics, ensuring your synthetic voice aligns with your content style without vocal fatigue during long sessions.
Best for Gaming Commentary
Felix is a strong choice for energetic promo-style content and fast-paced gaming updates. His bright, upbeat delivery maintains listener engagement during high-action segments.
- Felix: Energetic, promo-style delivery.
- Ideal for quick updates and hype moments.
Best for Storytelling
Luna offers a soft, storytelling tone suitable for narrative-driven streams. Her clear, warm narration reduces listener fatigue during longer, atmospheric content.
Nova provides a brighter alternative for upbeat, conversational storytelling where a slightly more active tone is preferred.
- Luna: Soft, storytelling focus.
- Nova: Bright, upbeat alternative.
Best for News Updates
Atlas delivers a neutral, professional tone suitable for clear news updates or factual reporting. His consistent pacing ensures information is conveyed without emotional bias.
Orion adds depth for documentary-style segments, offering a serious, authoritative presence suitable for detailed analysis or lore explanations.
- Atlas: Neutral, news-ready clarity.
- Orion: Deep, documentary authority.
Listen to the 8 voices
Each sample below was generated by this API from the same sentence. We check for listener fatigue, ensuring the voice remains pleasant and clear during hour-long streams without becoming grating.
23 languages
Any voice speaks any of these languages.
