Frequently asked questions
Are tts-1 and tts-1-hd different models?
They are two presets for the same underlying engine. Both are distinct model identifiers in our API, but they use the same underlying synthesis engine. The distinction is purely a label for your request body, allowing you to select your preferred output style without changing your integration logic.
Do they have different prices?
No. Both tts-1 and tts-1-hd share the same price point. You pay $10 per 1 million characters regardless of which preset you choose. The cost is determined by the input text length, not by the model variant selected in your API request.
Can I switch between them easily?
Yes, switching is immediate. Simply update the model field in your JSON payload from "tts-1" to "tts-1-hd" or vice versa. No code changes are required beyond this single field. Your existing integration continues to work seamlessly, allowing you to test both outputs with zero friction.
Which one sounds better?
"Better" depends on your use case. Both use the same 8 voice presets and engine. The choice is subjective based on audio characteristics you prefer. Since pricing and capabilities are identical, you can generate samples for both and decide which fits your project's audio profile best.
Is there a quality difference?
Both presets utilize the same core synthesis engine and voice library. The primary difference lies in the audio processing characteristics. Since they share the same underlying technology, the quality variance is minimal. You can test both to see which output aligns better with your specific audio requirements.
Do they use the same voices?
Yes. Both presets support all 8 voice presets: aria, nova, luna, sage, felix, milo, atlas, and orion. You can select any voice regardless of whether you choose tts-1 or tts-1-hd. The voice selection is independent of the model preset you specify in your request.
Does the choice affect generation speed?
No. Generation speed is consistent regardless of the model preset. Audio is generated at roughly real-time speed, approximately 25 characters per second. Long texts are split and joined automatically. Your request will queue similarly whether you use tts-1 or tts-1-hd under load.
Are there different rate limits for each?
No. Rate limits apply per API key, not per model. You are limited to 60 requests per minute (10 before top-up). This limit covers all requests to the speech endpoint, regardless of whether you are using tts-1, tts-1-hd, or mixing both in your application.
Which one is better for long texts?
Both handle long texts identically. Texts exceeding 4096 characters are split into parts and the resulting audio files are joined. Since the engine and pricing are the same, neither preset offers a technical advantage for long-form content. Choose based on your audio preference rather than text length.
Is one more suitable for mobile apps?
Both are equally suitable. The API returns standard audio formats like mp3, wav, opus, flac, and pcm. Since pricing, speed, and voice options are identical, your mobile app can use either preset without performance or cost penalties. Select the one that best matches your desired audio tone.
Do they support all languages?
Yes. Both presets support all 23 available languages, including English, Spanish, German, Japanese, and Arabic. Language support is determined by the voice and engine, not the model preset. You can specify the language parameter or rely on automatic detection for either tts-1 or tts-1-hd.
Can I use both in the same project?
Absolutely. You can use both presets within the same application. Since they share the same rate limits and pricing, you can dynamically switch between them based on user preference or content type. This allows you to offer variety without increasing your infrastructure complexity or costs.
Which one is more reliable?
Both share the same reliability profile. They process requests through the same pipeline and are subject to the same queue and rate limits. Failed requests are refunded in full for both. There is no difference in uptime or error handling between tts-1 and tts-1-hd.
Do I need to change my code to switch?
No. You only need to update the model field in your JSON body. The rest of your request structure, including input, voice, and response_format, remains unchanged. This OpenAI-compatible endpoint design ensures that switching between presets requires minimal code modification, typically just a string change.
Same engine, same price
Both models rely on the same underlying text-to-speech engine, meaning the core audio quality and voice characteristics remain consistent regardless of your choice. You are simply selecting between standard and high-definition processing paths within the same system architecture.
Both models use the same TTS engine
The underlying synthesis technology is identical for both tts-1 and tts-1-hd. This means the voice presets, language support, and tonal qualities are consistent across both options. You get the same 8 voices, from Aria to Orion, regardless of which model you select.
Pricing is identical per character
You pay exactly the same rate for both models: $10 per million characters processed. There is no premium charge for the HD option. Whether you generate a short test phrase or a long audiobook chapter, the cost per character remains constant at $0.01 per 1,000 characters.
No performance difference in cost
Since the pricing is identical, your billing logic does not need to differentiate between standard and high-definition requests. You only pay for the characters you actually use, and failed requests are free. This simplicity makes budgeting straightforward for indie projects without hidden tier differences.
Use cases for tts-1
Tts-1 is our standard model, designed for straightforward text-to-speech needs where clear, intelligible audio is the priority. It handles everyday narration, informational content, and general purpose voiceovers with consistent quality, making it a reliable baseline for most projects.
Standard quality audio
The engine delivers clear, natural-sounding speech suitable for most standard use cases. It strikes a balance between clarity and generation speed, ensuring your audio sounds professional without unnecessary complexity.
Faster generation for bulk content
Requests are processed sequentially. Since tts-1 and tts-1-hd use the same engine, generation speed is roughly the same for both. This is ideal for indie developers needing to produce many audio files quickly.
- Handle up to 4096 characters per request
- Efficient for high-volume batch processing
- Real-time speed on long texts
Ideal for prototypes and testing
Because there is no price difference between models, tts-1 is perfect for rapid iteration. You can generate audio for your MVP or test flows without worrying about costs. Switch to HD only when you are ready for final client deliverables.
Use cases for tts-1-hd
tts-1-hd is designed for scenarios where audio clarity is non-negotiable. While it uses the same underlying engine as tts-1, it delivers higher fidelity audio suitable for professional projects.
This model excels in producing high-fidelity audio that sounds polished and ready for public consumption, making it ideal for final deliverables rather than rough drafts.
Higher fidelity audio
The HD model captures more nuance in speech, providing a smoother listening experience. This is crucial when users will listen to the audio repeatedly.
Audio quality is enhanced, providing a smoother listening experience.
Better for final production
When you are preparing content for a podcast, audiobook, or commercial, tts-1-hd ensures the audio meets professional standards. It saves you from needing additional post-production polish.
Freelancers often switch to HD for client-facing projects to ensure the highest quality output. This minimizes the risk of clients requesting reworks due to audio quality.
Preferred for premium content
If your project involves premium subscribers or paid content, the enhanced audio quality justifies the experience. Listeners notice the difference in clarity and depth.
Use HD for marketing materials, educational courses, or any content where audio quality is a key selling point. It adds a layer of professionalism to your work.
Technical specifications
Both tts-1 and tts-1-hd share an identical request structure, making them interchangeable in your code. You send a JSON body with the same fields, and receive the same audio formats back.
Same input and output formats
The API accepts the same input text (up to 4096 characters) and language parameters for both models. You can choose from mp3, wav, opus, flac, or pcm for the response_format regardless of which model you pick.
- Input: Text string, max 4096 chars
- Output: mp3, wav, opus, flac, pcm
Same voice presets available
All eight voice presets—aria, nova, luna, sage, felix, milo, atlas, and orion—are available for both models. There are no voice restrictions based on the model choice, ensuring consistent character behavior across quality levels.
Same speed control options
Speed control works identically for both. You can adjust the speed parameter between 0.25 and 4.0 to slow down or speed up the narration. This gives you precise control over the final audio timing without needing separate configurations.
When to choose which
The model choice primarily affects audio fidelity rather than cost or integration. Both tts-1 and tts-1-hd use the same underlying engine, so you pay exactly $10 per million characters regardless of which you select. This means your budget remains predictable whether you prioritize raw output speed or higher resolution audio.
Choosing between them depends on your specific use case. For rapid prototyping, background narration in apps, or bulk content generation where file size and generation time matter, tts-1 is the efficient default. It delivers clear, standard quality without unnecessary overhead.
Choose tts-1 for speed and volume
Use tts-1 when you need to generate large volumes of audio quickly or when the audio serves as background elements. Since requests are processed sequentially, faster generation helps keep your queue moving during high-demand periods. It is ideal for prototypes, internal tools, or applications where ultra-high fidelity is not critical.
- Ideal for bulk content generation
- Faster generation for rapid iteration
- Standard quality sufficient for background audio
Choose tts-1-hd for quality
Select tts-1-hd for final production where audio quality is paramount. It is better suited for premium content, podcast intros, or direct-to-user interfaces where listeners might scrutinize the sound. While the price per character is identical, the higher fidelity makes it the preferred choice for client deliverables.
- Higher fidelity for final products
- Preferred for premium content
- Better for direct-to-user interfaces
Test both to see your preference
Since the integration is identical, you can easily swap models in your code to compare results. Generate a sample with both models and listen closely. If you cannot hear a significant difference in your specific use case, stick with tts-1 for simplicity. If the HD version adds necessary clarity, switch your production endpoint.
- Switch models in one field
- No code changes required
- Test both to confirm preference
Integration ease
Switching between models is as simple as changing a single string in your request body. You don't need to rewrite your integration logic or update your SDK configuration when you want to upgrade quality or cut costs.
Switch models in one field
Both models accept the same model field parameter in your JSON payload. Simply toggle the value between tts-1 and tts-1-hd to change the engine instantly. This makes A/B testing your audio output during development trivial.
No code changes required
Since the request shape is identical, your existing code handles both seamlessly. You don't need conditional logic or separate endpoints for different quality levels. Just update the model string and send the request.
Same API endpoint for both
Both models use the same POST /v1/audio/speech endpoint. Your base URL remains https://api.play35bd.com/v1, so your infrastructure stays simple. This consistency reduces debugging time and keeps your API client clean.
Audio file size
HD may produce larger files
The tts-1-hd model generates higher-fidelity audio, which often results in larger file sizes compared to the standard tts-1 output. This is particularly noticeable when using lossless formats like WAV, FLAC, or uncompressed PCM.
MP3 format mitigates size issues
Choosing the MP3 response format significantly reduces file size while maintaining good quality. This makes it a practical choice for most use cases, balancing audio clarity with efficient storage and transfer.
Consider bandwidth for HD
If you are processing large volumes of audio or transferring files to low-bandwidth environments, stick to MP3 or AAC. Reserve HD for final production where audio fidelity is critical, and ensure your infrastructure can handle the increased data load.

Community feedback
Many indie developers use tts-1-hd for final client deliverables where audio clarity matters most. Its higher fidelity ensures that the synthetic speech sounds polished and professional, meeting the quality standards expected in finished products.
tts-1 remains popular for rapid iteration. Developers use it to quickly generate audio for prototypes and testing phases, where speed and volume take precedence over perfect fidelity. This model allows for faster turnaround times during the creative process.
Indie devs prefer HD for final products
Many freelancers switch to tts-1-hd when preparing assets for release. The enhanced audio quality makes it the preferred option for premium content, ensuring that the final output meets high production standards without needing additional post-processing.
tts-1 used for rapid iteration
During the development phase, tts-1 is favored for its speed. It allows developers to test voice options and script variations quickly, making it ideal for bulk content generation or when working with tight deadlines on early drafts.
Both are widely used
Both models are widely adopted by the indie community. Developers often keep both available, using tts-1 for internal testing and tts-1-hd for client-facing work. This flexible approach ensures that projects move efficiently from concept to completion.
The choice ultimately depends on the specific needs of each project, with many teams maintaining a workflow that leverages the strengths of both models.
Pricing comparison
Both models charge exactly $10 per million characters. There is no price difference between the standard and HD options, so your budget remains predictable regardless of quality choice. Spaces and punctuation count as characters in this calculation.
Failed requests are free, so you only pay for successful synthesis. The billing logic is identical for both models, simplifying cost estimation for your projects.
$10 per million characters each
Each model charges $10 per million characters processed. This rate applies uniformly to both tts-1 and tts-1-hd, ensuring consistent costs across quality tiers.
No price difference
You pay the same rate for either model. Whether you generate audio for a quick prototype or a final client deliverable, the character cost remains unchanged.
Same billing logic
Billing treats both models identically. Failed requests are refunded, and successful generations are charged at $0.01 per 1,000 characters. Your balance updates automatically after top-ups via USTRC-20.
Voice compatibility
Both model variants support the full set of eight voice presets, from Aria’s warm narration to Orion’s deep documentary tone. You can switch between any voice regardless of whether you select tts-1 or tts-1-hd, giving you consistent creative control across all quality tiers.
There are no voice restrictions tied to specific models. Every voice behaves identically in terms of capabilities, allowing you to swap models without worrying about losing access to your preferred speaker.
All 8 voices work with both
Each voice preset, including Nova and Felix, is available for both standard and high-definition generation. You choose the voice based on your content needs, not your quality tier.
This means your library of characters remains intact whether you are generating quick prototypes or final client deliverables.
- Aria, Nova, Luna, and Sage are available in both models
- Felix, Milo, Atlas, and Orion work identically across both tiers
- No voice is locked to only one model
No voice restrictions per model
Unlike some platforms that limit premium voices to higher tiers, here all eight voices are fully accessible with either model choice. Your account key grants access to the entire voice library regardless of the model parameter.
This uniformity simplifies your integration logic, as you do not need conditional code to handle voice availability based on the selected model.
Consistent voice behavior
The synthesis engine processes voices consistently, ensuring that a voice sounds like itself whether generated in standard or HD mode. The primary difference remains audio fidelity, not voice identity or character traits.
This consistency helps maintain brand voice continuity across your projects, whether you are batch-processing large texts or crafting nuanced short phrases.
Latency differences
Generation speed depends on text length and server queue. Short phrases take a few seconds regardless of model choice. Long texts process at roughly real-time speed, meaning about 25 characters of text per second of audio generated.
Similar generation times
For most standard requests, tts-1 and tts-1-hd feel instant. The difference in processing time is negligible for short clips or bulk generation where parallelism helps.
- Short texts: ~2-5 seconds
HD may take slightly longer
The high-definition model requires more computation per character. You might notice a small delay compared to tts-1, but it remains within acceptable limits for most indie workflows.
Dependent on server load
Both models share the same processing queue. Under heavy load, either model may wait in line. The actual wait time depends on current traffic, not the model selected.
Recommendation for freelancers
For indie developers and freelancers, your workflow should reflect the stage of production. Start with tts-1 during the prototyping phase. It generates audio quickly, allowing you to iterate on script timing and voice selection without waiting long for each render. This speed is crucial when you are refining the narrative flow before committing to final assets.
Once the script is locked and ready for client delivery, switch to tts-1-hd. The higher fidelity ensures the final audio meets professional standards for premium content. Since both models share the same pricing structure, you only pay for the quality tier you need at that moment, keeping your costs transparent and predictable.
Start with tts-1 for testing
Use tts-1 for rapid iteration. It processes requests fast, letting you test multiple voices and speed settings in minutes. This is ideal for drafting voiceovers where minor quality differences don't impact the creative decision-making process.
You can generate bulk drafts to check pacing. Since the engine is the same, the timing estimates remain consistent, helping you plan video edits accurately before investing time in high-res renders.
- Ideal for internal reviews and script debugging
- Fast generation supports quick A/B testing of voices
- Perfect for rough cuts and storyboarding
Switch to HD for client deliverables
When the project moves to final production, use tts-1-hd. The improved audio fidelity is noticeable in the final mix, making it the preferred choice for customer-facing content. It ensures your deliverables sound polished and professional.
Since pricing is identical per character, there is no financial penalty for choosing higher quality. You simply switch the model field in your API request to secure the best sound for your end users.
Reserve tts-1-hd for the final render of key scenes or main narrations. This targeted approach lets you balance quality with efficiency, ensuring every dollar spent contributes to a premium listener experience.
- Ensures professional-grade audio for end users
- No extra cost compared to standard model
Use both as needed
Mix models based on project needs. Use tts-1 for background elements or minor lines where audio quality is less critical, and tts-1-hd for lead narration. This flexible approach maximizes your budget without compromising on key audio moments.
Keep your integration simple. Both models use the same API endpoint and request structure, so switching between them requires only changing the model parameter. This ease of use allows you to optimize your pipeline dynamically.
Test both voices to see which fits your brand voice best. Since all eight voices work with both models, you can maintain consistent character tones across different quality levels, ensuring a cohesive audio experience.
- Optimize costs by matching quality to importance
- Simple code changes switch models instantly
- Maintain consistent voice identity across projects
Listen to the 8 voices
Each sample below was generated by this API from the same sentence.
23 languages
Any voice speaks any of these languages.
Sources and further reading
- Wikipedia
- LibriVox public-domain audiobooks
- openai-node on GitHub
- openai-python on GitHub
- LibriTTS-R corpus (OpenSLR 141)
- FFmpeg documentation
- Creative Commons Attribution 4.0
- IETF RFC 7845: Ogg Opus
- curl documentation
- Node.js documentation
- OpenAI API reference: create speech
- W3C WAI: making audio and video accessible
