What is the Difference Between TTS-1 and TTS-1-HD?

By NodeVoice editorial team · Updated 2026-11-18

This page clarifies the technical and economic distinctions between the tts-1 and tts-1-hd models within the NodeVoice API. Both models utilize the same synthesis engine and voice presets, ensuring identical audio character quality and processing latency. The primary difference lies in the output format capabilities, allowing developers to select the optimal balance between file size and fidelity for their IoT applications.

Readers will find precise guidance on choosing between these models based on bandwidth constraints and device processing loads. Since both options cost $0.01 per 1,000 characters, this guide helps indie developers optimize their smart home audio infrastructure without unnecessary expenditure, ensuring deterministic pricing for edge computing deployments.

What is the Difference Between TTS-1 and TTS-1-HD?

Frequently asked questions

Do both models support all 23 languages?

Yes. Both tts-1 and tts-1-hd use the same engine and support all 23 listed languages. Any voice preset can speak any of these languages. If the language is not specified, the system guesses it from the script, defaulting to English for Latin text.

Is there a speed difference between the two models?

No. Both models share the same engine and pricing. Generation speed is approximately real-time, processing about 25 characters of text per second. Performance depends on queue load rather than the model selected, as requests are processed sequentially.

Can I switch models mid-stream?

This API does not support streaming or WebSocket connections. Each request is a discrete POST operation returning a complete audio file. You must select the model (tts-1 or tts-1-hd) for each individual request when sending text to the endpoint.

Do both models produce watermarked audio?

Yes. Every generated audio file, regardless of whether it uses tts-1 or tts-1-hd, carries an inaudible watermark. This marks the file as synthetic speech, ensuring consistent metadata and provenance tracking for both standard and high-definition outputs.

Which model is better for smart speaker alarms?

Both models are identical in engine and price, making them equally suitable for alarms. For IoT devices, tts-1 is often preferred for lower latency if queue wait times are critical, though both provide the same audio quality and voice presets.

Is the HD model harder on CPU resources?

The API is cloud-based, so CPU processing occurs on our servers, not your device. Both models use the same engine. Your hardware only handles the API request and receives the resulting audio file, with no difference in client-side resource usage.

Do I pay more for the HD model?

No. Both tts-1 and tts-1-hd have the same engine and price. You pay $10 per 1 million characters for either model. Failed requests are free, and the cost per character remains identical regardless of which model you select.

Are there any voice presets exclusive to HD?

No. All eight voice presets (aria, nova, luna, sage, felix, milo, atlas, orion) are available for both models. There is no difference in voice selection between tts-1 and tts-1-hd; the distinction is purely in the model identifier used for the request.

How does the language detection work for both?

Detection works identically for both models. If the language field is omitted, the system guesses the language from the script. Latin script defaults to English. For non-English Latin text, you should explicitly set the language parameter to ensure correct synthesis.

Can I use the same API key for both models?

Yes. A single API key works for both tts-1 and tts-1-hd. You simply change the model field in your JSON body to switch between them. The key supports up to 60 requests per minute, regardless of which model is used.

What happens if the HD model fails?

If synthesis fails for either model, the charge is returned in full. The billing system does not distinguish between tts-1 and tts-1-hd for refunds. Your balance remains unaffected, and you can retry the request with the same API key.

Is the output format different for HD?

No. Both models support the same response formats: mp3, wav, opus, flac, and pcm. The output format is determined by your request parameters, not the model choice. Both models deliver audio files with the same structural properties.

Do both models support the same speed range?

Yes. Both models support a speed range from 0.25 to 4.0. The speed parameter applies equally to tts-1 and tts-1-hd. You can adjust the playback speed for any voice preset without any difference in behavior between the two models.

Which model has lower latency in queues?

Latency is similar for both, as they share the same engine and queue processing. Requests are processed one after another. Under load, both models may experience delays. There is no priority queue difference between tts-1 and tts-1-hd.

Model Architecture Overview

Sound wave illustration for: Model Architecture Overview

NodeVoice utilizes a single underlying synthesis engine for both tts-1 and tts-1-hd models, ensuring deterministic behavior across IoT deployments. The distinction lies in the audio fidelity of the output. This architecture allows embedded devices to switch modes without requiring separate voice pipelines or retraining of local inference modules.

Both models support the same eight voice presets, maintaining consistent timbre and articulation regardless of the selected quality tier. This uniformity simplifies integration for smart home systems where voice consistency is critical for user experience across different hardware tiers.

Synthesis Engine Comparison

The tts-1 model is suitable for real-time alerts. The tts-1-hd model provides higher audio fidelity. Both endpoints accept identical JSON bodies, ensuring seamless transition between standard and high-definition outputs.

  • Single neural network backbone
  • Identical voice preset availability
  • Same request payload structure

Audio Quality Metrics

Audio quality differences are primarily perceptual, with HD providing higher fidelity. For IoT applications, this enhances speech intelligibility in noisy environments. The inaudible watermark remains present in both formats, marking all outputs as synthetic speech for compliance tracking.

  • Higher audio fidelity in HD
  • Improved audio quality
  • Consistent voice timbre

Processing Latency Differences

Processing latency is generally comparable, with both models processing requests sequentially. Long texts may experience queuing under load, regardless of the model selected. Processing latency is generally comparable for both models.

  • Sequential request processing
  • Negligible latency difference
  • Queue-dependent delays

Pricing Structure

Sound wave illustration for: Pricing Structure

Both tts-1 and tts-1-hd utilize the identical synthesis engine and are billed at the same standard rate of $10 per 1 million characters. This uniform pricing structure eliminates financial ambiguity for IoT developers, ensuring that budget forecasts remain stable regardless of the selected audio fidelity tier.

Identical Cost Per Character

The cost metric is strictly defined by character count, including spaces and punctuation. Whether selecting the standard tts-1 model or the high-definition variant, the charge remains $0.01 per 1,000 characters. This consistency simplifies cost-per-request calculations for embedded devices generating voice responses.

  • Rate: $10 per 1,000,000 characters
  • Includes spaces and punctuation
  • Failed requests are free

No Premium for HD

Users do not pay extra for the TTS-1-HD model. The tts-1-hd option provides higher audio fidelity without a price premium, allowing developers to upgrade audio quality for critical alerts or narrations without increasing infrastructure costs.

Budget Planning for IoT

With deterministic pricing, developers can accurately predict expenses for smart home devices. The absence of tiered pricing prevents unexpected costs during peak usage, supporting reliable operational budgeting for large-scale IoT deployments.

  • Predictable per-character billing
  • No hidden tier fees
  • Supports scalable IoT budgets

Use Case Recommendations

Sound wave illustration for: Use Case Recommendations

NodeVoice’s TTS-1 and TTS-1-HD models share identical pricing ($10 per 1 million characters) and engine performance, making the choice purely a matter of audio fidelity requirements for your IoT deployment.

Standard Quality for Alerts

Use TTS-1 for time-sensitive notifications where clarity outweighs richness. Voices like atlas (neutral, news) or felix (energetic, promo) deliver crisp alerts efficiently. The standard model processes text at roughly real-time speed, ensuring minimal delay for smart home triggers.

High Definition for Narration

Select TTS-1-HD for longer-form audio such as audiobook readings or detailed instructions. Voices like luna (soft, storytelling) or orion (deep, documentary) provide higher fidelity. This ensures natural intonation for edge computing devices handling narrative content.

Bandwidth Considerations

Both models support mp3, wav, opus, flac, and pcm. For constrained IoT networks, use opus or mp3 to reduce data transfer. WAV and FLAC offer lossless quality but increase bandwidth usage. PCM provides raw data for immediate device playback.

  • Opus: Best balance of quality and size.
  • MP3: Widely supported, efficient compression.
  • WAV/FLAC: Higher bandwidth, lossless quality.

Technical Implementation

Sound wave illustration for: Technical Implementation

Both tts-1 and tts-1-hd accept identical request parameters, allowing seamless interchangeability within your IoT integration logic. The API endpoint remains constant, requiring only a model string change to switch synthesis quality without altering the JSON structure or authentication headers.

Parameter Compatibility

Voice selection and speed settings (0.25–4.0) apply uniformly across both models. You may use any of the eight preset voices with either engine, ensuring consistent voice identity regardless of the chosen audio fidelity level for your smart home devices.

Voice Selection Consistency

Each voice preset maintains its distinct tonal characteristics in both standard and high-definition modes. This consistency allows developers to assign specific voices to different hardware interfaces without retraining or adjusting voice profiles when upgrading audio quality.

Format Support

The API supports mp3, wav, opus, flac, and pcm output formats for both models. Select opus or mp3 for bandwidth-constrained edge devices, or wav for uncompressed audio processing in local smart home hubs.

Performance in Low Latency Environments

Sound wave illustration for: Performance in Low Latency Environments

NodeVoice’s TTS API processes requests sequentially, meaning latency depends on queue depth. Short phrases require a few seconds, while longer texts generate at roughly real-time speed (25 characters per second). Under load, requests may wait in a queue before synthesis begins.

Both tts-1 and tts-1-hd share the same engine and pricing, ensuring consistent latency behavior. Developers should account for this sequential processing when designing IoT devices that require immediate voice feedback, as simultaneous requests will experience increased wait times proportional to system load.

Response Time Analysis

Generation time scales with input length. A short phrase takes a few seconds, whereas long texts approach real-time synthesis speed. Requests are processed one after another, so high concurrency increases delay.

For IoT alerts, keep inputs under 4096 characters to minimize queue wait. If latency is critical, consider pre-generating common phrases during idle periods rather than on-demand synthesis.

  • Short phrases: few seconds generation time
  • Long texts: ~25 characters per second synthesis speed
  • Sequential processing causes queue delays under load

Network Impact

Audio formats vary in size. MP3 offers small payloads for constrained networks, while WAV and FLAC are larger. Opus provides efficient compression for low-bandwidth IoT connections.

Select formats based on your device’s network capability. Smaller files reduce transmission time, but ensure the decoder supports the chosen format. PCM is uncompressed and largest, suitable only for local playback.

  • MP3: optimal for low-bandwidth IoT networks
  • WAV/FLAC: larger payloads, lossless quality
  • Opus: efficient compression for constrained devices

Device Processing Load

Decoding audio formats requires CPU resources. MP3 and Opus are efficient, while PCM and WAV demand more processing power. Ensure your IoT hardware can handle real-time decoding without blocking other tasks.

Test decoding performance on target hardware. High CPU usage during playback may affect other sensor readings or network operations. Use lightweight decoders if processing load is a concern.

  • MP3/Opus: low CPU decoding overhead
  • WAV/PCM: higher processing requirements
  • Monitor CPU usage during playback to avoid blocking

Audio Format Optimization

Sound wave illustration for: Audio Format Optimization

The selected response_format directly impacts bandwidth usage and device storage for embedded IoT nodes. Choosing between lossy and lossless codecs requires balancing audio fidelity against transmission efficiency in constrained networks.

MP3 vs WAV Size

WAV files provide uncompressed audio but consume significantly more bandwidth. For a standard 10-second alert, a WAV file may exceed several megabytes, whereas MP3 reduces this to a fraction. This compression is critical for devices with limited data plans or slow connections.

  • WAV: Uncompressed, high file size, instant playback.
  • MP3: Compressed, small file size, standard compatibility.

Opus Efficiency

Opus offers superior compression ratios with minimal quality loss, making it ideal for real-time voice responses. It maintains clarity even at low bitrates, ensuring that voice commands or notifications remain intelligible on low-bandwidth IoT networks without excessive data consumption.

  • Optimal for low-bitrate streaming.
  • High efficiency for variable data rates.

FLAC Lossless Benefits

FLAC provides lossless compression, preserving original audio quality while reducing file size compared to WAV. This format is suitable for archival purposes or high-fidelity narration where bit-perfect reproduction is required, though it requires more storage and processing power on the device.

  • Lossless compression preserves quality.
  • Better than WAV for storage efficiency.

Developer Experience

Sound wave illustration for: Developer Experience

Switching models requires only changing the model field in the JSON payload between "tts-1" and "tts-1-hd". Both use the same endpoint and parameters, ensuring consistent integration within your IoT firmware or smart home hub without code refactoring.

Testing involves sending short text samples via POST /v1/audio/speech and verifying the returned audio files. Use the provided curl example to confirm latency and format compliance before deploying to edge devices.

Switching Between Models

Switching models requires only changing the model field in the JSON payload between "tts-1" and "tts-1-hd". Both use the same endpoint and parameters, ensuring consistent integration within your IoT firmware or smart home hub without code refactoring.

Both models support the same eight voices and all output formats, allowing seamless fallback strategies if one model experiences temporary queue delays.

  • Identical request structure for both models
  • Same voice presets available in both tiers
  • No API endpoint changes required during switch

Testing Protocols

Testing involves sending short text samples via POST /v1/audio/speech and verifying the returned audio files. Use the provided curl example to confirm latency and format compliance before deploying to edge devices.

Verify that the response_format matches your device’s decoder capabilities, such as MP3 for general use or WAV for uncompressed audio needs in critical alerts.

  • Validate audio playback on target hardware
  • Check character count limits (4096 chars)
  • Confirm authentication header format

Debugging Audio Artifacts

Audio artifacts typically stem from network timeouts or truncated text inputs. Ensure your client handles HTTP errors correctly, as failed requests return no audio data.

Use the language parameter explicitly if the script is ambiguous to prevent incorrect voice synthesis, which may sound like an artifact to end-users.

  • Check HTTP status codes for errors
  • Verify text length does not exceed 4096 characters
  • Set explicit language codes for Latin-script non-English text
Developer Experience

Future Compatibility

Sound wave illustration for: Future Compatibility

The API maintains backward compatibility for tts-1 and tts-1-hd endpoints, ensuring existing IoT integrations remain functional without immediate code changes. Deprecation of either model will follow a standard notification period, allowing time to adjust voice parameters or switch models in embedded devices.

Model Deprecation Plans

We do not guarantee indefinite permanence for specific model versions. If a model is deprecated, we will provide advance notice via the API status page, allowing developers to update their firmware or configuration files before the old model becomes unavailable.

SDK Support Longevity

Since we use the OpenAI-compatible endpoint structure, your existing SDK implementations (e.g., Python, JavaScript) will continue to work as long as the base URL remains valid. We do not maintain a proprietary SDK, reducing long-term maintenance overhead for your team.

Standardization Trends

Adhering to the OpenAI speech API shape ensures your code remains portable. This standardization reduces vendor lock-in, allowing you to switch providers if needed while keeping the same request format and response handling logic for your smart home devices.

Community and Support

Sound wave illustration for: Community and Support

Support for NodeVoice operates exclusively via email. There is no live chat, phone line, or in-app messaging. Developers should expect standard response times for technical inquiries regarding API integration or billing discrepancies.

Unlike enterprise platforms, we do not provide real-time status pages or SLA guarantees. All communication is asynchronous, allowing your team to document issues thoroughly before submission.

Common Questions

The most frequent question concerns the character limit per request. The API enforces a strict 4096-character maximum for the input field. Requests exceeding this limit are rejected with a 400 error.

Another common query relates to voice language detection. If the language field is omitted, the system guesses based on script. Latin script defaults to English; non-English Latin text requires explicit language codes.

  • Request limit: 60 requests per minute per key.
  • Character limit: 4096 characters per request.
  • Auth header: Authorization: Bearer tts-...

Error Logs Analysis

Error responses include standard HTTP status codes. A 429 indicates rate limit exceeded (60 req/min). A 400 signals invalid parameters, often due to unsupported formats or invalid voice IDs.

Check the error object in the JSON response for details. Failed synthesis returns a full refund; no charge is applied. Ensure your client handles retries with exponential backoff.

  • 400: Invalid request parameters.
  • 401: Invalid or expired API key.
  • 429: Rate limit exceeded.

Best Practices

Always split long texts into 4096-character chunks before sending. Join the resulting audio files client-side to maintain continuity. This avoids single-request failures.

Use the speed parameter to adjust playback without re-generating audio. This reduces API calls and saves credits. Monitor your balance regularly via the dashboard.

  • Split text at <4096 chars.
  • Use speed for playback adjustment.
  • Monitor balance in /get-api-key/.

Summary of Differences

Sound wave illustration for: Summary of Differences

This document confirms that tts-1 and tts-1-hd utilize the same synthesis engine with identical pricing at $0.01 per 1,000 characters. The distinction lies solely in output fidelity, affecting storage and bandwidth requirements for IoT devices.

Developers should select the model based on audio quality needs versus network constraints. Both models support all eight voices and standard formats, ensuring consistent voice behavior across deployment environments.

Key Distinctions

The primary difference is audio fidelity. tts-1-hd provides higher resolution audio, while tts-1 offers standard quality. Both models share the same character limit, speed range, and voice presets.

  • Same engine and pricing

Selection Criteria

Choose tts-1 for simple alerts where bandwidth is limited. Select tts-1-hd for narration or detailed instructions where clarity is critical. Both support all 23 languages and formats.

Final Recommendation

For most smart home applications, tts-1 provides sufficient quality. Use tts-1-hd only when high-fidelity audio is explicitly required by the use case.

Additional Resources

Sound wave illustration for: Additional Resources

Access the comprehensive API reference to understand endpoint structures and parameter limits. The documentation details the exact JSON schema required for the speech endpoint, ensuring seamless integration into smart home hubs and IoT edge devices without manual formatting errors.

Explore practical implementation guides that demonstrate how to handle rate limits and split long texts. These resources help developers optimize character usage within the 4096-character constraint, reducing latency for real-time voice responses in embedded systems.

Documentation Links

The official API documentation provides complete specifications for the tts-1 and tts-1-hd models. It includes reference tables for all 8 voices, supported languages, and output formats like MP3, WAV, and Opus. This ensures deterministic integration for hardware manufacturers.

Use the GET /v1/voices endpoint to retrieve available voices programmatically. This allows your application to dynamically select the appropriate voice based on the device type or user preference without hardcoding values.

  • API Reference: https://play29p.com/docs/
  • Voice List Endpoint: GET /v1/voices
  • Rate Limit Specifications: 60 requests/minute

Code Examples

Review ready-to-use curl commands for generating speech with the tts-1 model. These examples demonstrate the exact JSON body structure, including the input, voice, and response_format fields required for successful synthesis.

See how to implement the OpenAI SDK with a custom base_url pointing to https://api.play29p.com/v1. This allows you to use standard libraries while maintaining compatibility with our specific endpoint structure.

  • cURL example for POST /v1/audio/speech
  • Python SDK snippet with base_url override
  • Node.js implementation for edge devices

Community Forums

Join the developer discussion to share integration experiences and troubleshoot common issues. The forum serves as a resource for indie developers working on smart home applications using NodeVoice's API.

Contribute to best practices for handling audio artifacts and optimizing bandwidth. Discuss strategies for splitting large texts and managing rate limits in high-throughput IoT environments.

Contact the Editorial Team

Sound wave illustration for: Contact the Editorial Team

NodeVoice relies on email for all communication, with no phone support or live chat available. Our editorial team handles inquiries directly, ensuring that every message is reviewed by personnel familiar with the underlying synthesis engine and API specifications.

Response times may vary depending on the volume of incoming requests. For urgent technical issues, include your API key and request ID in the subject line to expedite processing.

Email Support

Direct all general questions to [email protected]. Include your account email and a clear description of the issue. We do not use automated ticketing systems; emails are read and answered by human staff.

Technical Inquiries

For API-specific questions regarding model behavior, voice parameters, or rate limits, use [email protected]. Provide the exact endpoint URL, request payload, and response headers. Include the request ID from the response header for precise log analysis.

Reference the specific model (tts-1 or tts-1-hd) and the voice preset used. This allows our team to replicate the synthesis conditions accurately.

  • Include request ID for log tracing
  • Specify model version and voice preset
  • Attach raw response headers if possible

Feedback Submission

Share your observations on voice quality or latency improvements via [email protected]. While we do not offer custom voice cloning, your data helps refine the existing preset behaviors.

Note that feedback does not guarantee immediate implementation. All inputs are reviewed periodically by the development team to align with the deterministic pricing and standard-compliant endpoint strategy.

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. This creates a more consistent user experience across different devices within a smart home or industrial network without manual reconfiguration.

Avatar of the voice aria
ariafemale
Avatar of the voice nova
novafemale
Avatar of the voice luna
lunafemale
Avatar of the voice sage
sagefemale
Avatar of the voice felix
felixmale
Avatar of the voice milo
milomale
Avatar of the voice atlas
atlasmale
Avatar of the voice orion
orionmale

23 languages

Any voice speaks any of these languages.

Sources and further reading

  1. Wikipedia
  2. W3C WAI: making audio and video accessible
  3. ISO 639 language codes
  4. openai-python on GitHub
  5. LibriTTS-R paper (arXiv 2305.18802)
  6. Creative Commons Attribution 4.0
  7. Tether (USDT)
  8. Node.js documentation
  9. openai-node on GitHub
  10. IETF RFC 7845: Ogg Opus
  11. Python documentation
  12. LibriVox public-domain audiobooks

More on this site