EchoCast Guide: Choosing the Right AI Voice for Your Show

By EchoCast editorial team · Updated 2026-11-18

This guide breaks down the eight available voice presets and audio formats for podcasters who prefer code over drag-and-drop tools. We map each synthetic voice—based on licensed recordings—to specific production workflows, from warm narration to neutral news delivery. This helps you select the right tone for episodic content, ensuring consistent branding across episodes without manual editing overhead.

Designed for developers integrating the EchoCast API, this page covers technical constraints like the 4096-character limit and speed ranges, alongside practical tips for batch processing. We also clarify licensing details, cost structures, and language support, so you can deploy broadcast-ready synthetic audio with confidence and precision.

EchoCast Guide: Choosing the Right AI Voice for Your Show

Frequently asked questions

Which voice is best for true-crime podcasts?

Orion’s deep, documentary-style tone suits true-crime well, while Atlas offers neutral, news-style narration. Neither is objectively best; choose based on your narrative style. Both voices are clear and consistent for episodic content. You can adjust the speed between 0.25 and 4.0 to match your specific pacing preferences.

Can I use Luna for children's podcasts?

Yes, Luna’s soft, storytelling voice suits children’s content. It supports all 23 languages and generates MP3, WAV, OPUS, FLAC, or PCM files. Luna is based on licensed LibriTTS-R recordings (CC BY 4.0) from LibriVox public-domain audiobooks. Each file carries an inaudible watermark marking it as synthetic speech.

Is Felix suitable for advertisement segments?

Yes. Felix is described as energetic and promo-focused, making it suitable for advertisements. You can adjust speed up to 4.0 for faster delivery. Each request supports up to 4096 characters, so split longer ad scripts into multiple requests.

Do all voices support all 23 languages?

Yes. All eight voices can speak all 23 supported languages. You can specify the language using an ISO code. If omitted, the system guesses from the script, defaulting to English for Latin text. This ensures consistent audio generation across your podcast episodes regardless of the voice preset you choose.

How does Aria differ from Luna?

Aria is a warm, clear narrator, while Luna is soft and storytelling-focused. Both are female voices supporting all 23 languages with speed adjustments from 0.25 to 4.0. Aria suits general narration, whereas Luna is better for immersive stories. Both are based on licensed recordings and carry an inaudible watermark marking them as synthetic speech.

Can I adjust the speed of Orion?

Yes. The speed parameter accepts values from 0.25 to 4.0 for any voice, including Orion. This allows you to control pacing for documentary-style content. Orion’s deep tone remains consistent across different speed settings. Note that requests are processed sequentially, so under load, generation may queue.

Is Atlas good for interview simulations?

Atlas delivers a neutral, news-style voice with clear, consistent delivery, making it suitable for interview simulations. You can adjust speed (0.25–4.0) and specify language. Atlas supports all 23 languages, allowing for versatile dialogue generation in various linguistic contexts.

Do the voices sound robotic?

The voices are synthetic, built on licensed recordings from LibriTTS-R and LibriVox. They are designed for natural flow and consistent quality, making them suitable for broadcast-ready audio. Each generated file includes an inaudible watermark to mark it as synthetic speech. While they avoid robotic artifacts by using real human voice data as a base, they remain clearly synthetic.

Can I clone a specific voice?

No, custom voice cloning is not offered. You cannot upload your own voice or design a new one. The eight presets (Aria, Nova, Luna, Sage, Felix, Milo, Atlas, Orion) are fixed. You can adjust speed and language, but not timbre.

How do I choose between MP3 and WAV?

Choose MP3 for smaller file sizes and broad compatibility. Choose WAV for lossless quality. Both are supported via the response_format parameter. MP3 is efficient for streaming, while WAV preserves full audio fidelity. Other formats like OPUS and FLAC are also available.

Are the voices based on real actors?

Yes. Our voices are based on licensed recordings from the LibriTTS-R dataset (CC BY 4.0) and LibriVox public-domain audiobook readings. They are synthetic interpretations of real human speech, not direct copies of specific actors. Each generated audio file carries an inaudible watermark to clearly identify it as synthetic speech.

Can I use multiple voices in one episode?

Yes. Generate audio for each voice separately via the API, then join the files to create your episode. Each request supports up to 4096 characters. You can assign any of the eight preset voices to different segments, allowing you to mix voices across your script as needed.

How much does it cost to generate 1 hour of audio?

Cost depends on character count, not duration. Pricing is $10 per 1 million characters. Since generation takes roughly real-time, 1 hour of audio requires approximately 100,000 characters. Failed requests are free, so you only pay for successful synthesis. Calculate your script’s character count to estimate the exact cost.

Do I need to credit the voices?

The voices are based on LibriTTS-R (CC BY 4.0). While attribution is recommended, our API does not enforce it. Each generated audio file contains an inaudible watermark marking it as synthetic speech to aid identification. We do not explicitly mandate specific credit formats for end-users.

Narrative and Storytelling

Sound wave illustration for: Narrative and Storytelling

EchoCast provides eight voice presets, with Aria, Luna, and Sage optimized for narrative structures. These female voices handle long-form scripts efficiently, delivering consistent audio for episodic content pipelines without requiring manual tone adjustments.

Aria: Warm Narrator

Aria offers a clear, warm tone ideal for standard narration. Her delivery remains steady across diverse scripts, making her suitable for general podcast segments and audiobook chapters.

Luna: Soft Storyteller

Luna provides a softer, more intimate delivery for storytelling. She excels in fiction and narrative-driven shows where emotional nuance enhances listener engagement.

Sage: Calm Instructor

Sage offers a calm, instructional tone perfect for educational segments. Her clarity ensures complex information is conveyed accurately, supporting informative content with minimal listener fatigue.

News and Journalism

Sound wave illustration for: News and Journalism

The API provides eight distinct voices, but three are particularly suited for news and journalistic formats. Atlas delivers a neutral, clear tone ideal for straight reporting where personality should not distract from the script. Felix offers high energy, making it effective for promotional segments or dynamic intros within a podcast episode. Milo provides a friendly, conversational voice that works well for informal updates or listener interactions.

While these voices handle standard news structures well, they are not exclusive to broadcast journalism. The engine supports 23 languages, allowing these presets to be used in various linguistic contexts. Adjusting the speed parameter (0.25–4.0) further tailors the delivery pace to match the specific rhythm of your show’s segments.

Atlas: Neutral News

Atlas is designed for clarity and neutrality, making it the standard choice for reading scripts where the content matters more than the voice personality. Its consistent tone ensures that listeners focus on the information rather than the delivery style. This makes it particularly effective for daily news updates or factual reporting where bias or emotional inflection might be undesirable.

  • Neutral, news-oriented tone
  • Ideal for straight reporting
  • Consistent delivery pace

Felix: Energetic Promo

Felix brings high energy to shorter segments, trailers, or call-to-action segments within a podcast. His energetic delivery helps capture attention quickly, making him suitable for promotional content or segment transitions. This voice type adds a dynamic element to your audio without requiring additional production work, keeping listeners engaged during shorter, high-impact sections.

  • Energetic, promo-focused delivery
  • Suitable for trailers
  • High engagement potential

Milo: Conversational

Milo provides a friendly, conversational tone that works well for interview intros, listener mail segments, or informal updates. His approachable voice creates a sense of connection with the audience, making complex topics feel more accessible. This voice is particularly effective for segments that require a more personal touch or a relaxed atmosphere.

  • Friendly, conversational tone
  • Good for listener mail
  • Approachable delivery
News and Journalism

Documentary and Deep Dives

Sound wave illustration for: Documentary and Deep Dives

For episodic deep dives, Orion provides a deep, documentary-grade voice that maintains authority over long scripts. Its consistent timbre ensures listeners remain engaged without distraction from vocal variations, making it suitable for serialized content where narrative continuity is critical.

The engine processes requests sequentially, so longer texts may experience slight queue delays during peak load. Despite this, Orion’s stable delivery supports complex storytelling structures without requiring manual pacing adjustments.

Orion: Deep Documentary

Orion delivers a deep, documentary voice ideal for long-form narrative content. Its consistent timbre ensures listeners remain engaged without distraction from vocal variations, making it suitable for serialized content where narrative continuity is critical.

The voice supports all 23 languages, allowing for authentic regional accents in historical or global topics. This flexibility ensures accurate pronunciation across diverse scripts without needing separate voice selections.

  • - Deep, authoritative tone for historical narratives
  • - Supports all 23 languages with accurate pronunciation
  • - Ideal for serialized episodic content

Voice Consistency

Each generated file carries an inaudible watermark marking it as synthetic speech, ensuring clear identification of AI-generated content. This watermark persists across all formats, including MP3 and WAV.

Consistency is maintained through the fixed 'tts-1' and 'tts-1-hd' models. Using the same voice and model settings across episodes ensures a uniform auditory brand identity for your podcast.

  • - Inaudible watermark in all outputs
  • - Fixed model ensures uniform tone
  • - No drift in voice characteristics over time

Tone Matching

To match tone, select Orion for serious, investigative pieces or Felix for energetic, promotional segments within your show. Avoid using Luna or Aria for these formats, as their softer tones may dilute the intended gravity.

Speed adjustments between 0.25–4.0 allow fine-tuning of pacing. For dense historical data, a slightly slower speed (0.8–1.0) enhances clarity without sacrificing the documentary feel.

  • - Pair Orion with serious narratives
  • - Use Felix for energetic promos
  • - Adjust speed 0.8–1.0 for clarity

Language Compatibility

Sound wave illustration for: Language Compatibility

EchoCast supports 23 languages, including Arabic, Chinese, French, German, Hindi, Japanese, Korean, Russian, Spanish, and Swahili. Any of the eight voice presets can speak any of these supported languages, providing flexibility for diverse podcast content.

Native Language Support

Each supported language is fully integrated into our engine. You can select any voice preset regardless of its original language, ensuring consistent audio quality across all 23 options.

Cross-Language Performance

Our engine delivers consistent results across all supported languages. Each voice maintains its unique tonal characteristics, whether speaking English, Japanese, or Spanish, ensuring a cohesive brand sound.

Language Detection Limits

If the language parameter is omitted, the system guesses the language from the script’s character set, defaulting to English for Latin-script text. For non-English Latin-script text, explicitly set the language parameter.

  • Languages not in the supported list, such as Vietnamese, Indonesian, Thai, Ukrainian, Romanian, Czech, and Bengali, are not processed.

Audio Format Selection

Sound wave illustration for: Audio Format Selection

Selecting the correct audio format is a critical decision for podcasters managing episodic content pipelines. The API allows you to specify response_format as mp3, wav, opus, flac, or pcm directly in the JSON request body, ensuring the output matches your specific distribution or editing workflow needs.

Each format serves a distinct purpose in the production chain, from final web delivery to raw archival storage. Understanding these differences helps maintain consistent audio quality while optimizing storage and bandwidth costs for your show's backend infrastructure.

MP3 for Web Distribution

MP3 is the standard for web distribution due to its balance of quality and file size. It ensures fast loading times for listeners on various devices without significant perceptible loss in audio fidelity for standard podcast consumption.

This format is ideal for final episode delivery, keeping bandwidth costs low while maintaining clear speech intelligibility across all 23 supported languages.

  • Optimal for streaming and mobile downloads
  • Smallest file size among lossy options
  • Universal player compatibility

WAV for Editing

WAV provides uncompressed audio, making it the preferred choice for post-production editing. It preserves the full dynamic range generated by the TTS engine, allowing for precise cuts, fades, and effects without generational loss.

Use this format when sending audio to editors or when further processing is required before final export.

  • Uncompressed, lossless quality
  • Ideal for DAW integration
  • Larger file size than MP3

FLAC for Archival

FLAC offers lossless compression, reducing file size by approximately 50-60% compared to WAV while retaining all audio data. This makes it excellent for long-term archival storage where storage space is a concern but quality must remain pristine.

It ensures your original generated voices remain available for future remastering or format updates without any degradation.

  • Lossless compression
  • Preserves all audio data
  • Smaller than WAV, larger than MP3

Technical Constraints

Sound wave illustration for: Technical Constraints

Technical constraints define the boundaries of your audio generation pipeline. Understanding these limits ensures reliable delivery of broadcast-ready audio without unexpected failures or delays.

The API processes requests sequentially, meaning under heavy load, synthesis may wait in a queue. Typical generation speed is roughly real-time, producing about 25 characters of text per second for long texts, though short phrases may take a few seconds.

Character Limits

Each request accepts up to 4096 characters, including spaces and punctuation. For longer scripts, split the text into parts within this limit and join the resulting audio files.

If the input exceeds 4096 characters, the request will fail. Always validate text length before sending to the API.

  • Max 4096 characters per request
  • Spaces and punctuation count toward the limit
  • Split longer texts manually before sending

Speed Ranges

Adjust playback speed between 0.25x and 4.0x to fit your pacing needs. Slower speeds improve clarity for dense narration, while faster speeds suit energetic promos.

Speed adjustments do not affect the character count or pricing. They only alter the audio output duration.

  • Range: 0.25 to 4.0
  • Does not affect pricing
  • Use for pacing control

Request Queues

Requests are processed one after another. Under load, a request may wait in a queue before synthesis begins.

This sequential processing ensures consistent engine performance but may introduce delays during peak usage times.

  • Sequential processing
  • Queues form under load
  • No parallel processing guarantee

Licensing and Rights

Sound wave illustration for: Licensing and Rights

EchoCast voices originate from the LibriTTS-R dataset (CC BY 4.0), derived from LibriVox public-domain audiobook readings. This foundation provides a stable, licensed base for your generated audio without complex per-voice licensing overheads.

LibriTTS-R Source

The synthesis engine uses recordings licensed under CC BY 4.0. While the source material is public-domain, the specific dataset licensing applies to the voice model structure rather than individual performer contracts.

  • Based on LibriTTS-R dataset
  • CC BY 4.0 licensed source
  • Derived from LibriVox readings

Watermark Identification

Every generated audio file contains an inaudible watermark, clearly marking the output as synthetic speech. This ensures transparency and helps identify AI-generated content in your podcast episodes or distributed media.

Commercial Use

You retain full rights to the generated audio files. The voices are not tied to specific identifiable individuals with separate contractual rights, simplifying commercial deployment for episodic content pipelines without additional attribution management.

Licensing and Rights

Integration Tips

Sound wave illustration for: Integration Tips

Integrate EchoCast into your podcast pipeline using standard HTTP requests or the OpenAI Python/Node SDKs. Simply update the base URL to https://api.playmm33.com/v1 and use your API key. This approach leverages existing client libraries without requiring custom code.

Authentication requires the Authorization: Bearer tts-... header. Since failed synthesis requests are free, you can safely retry transient errors without incurring additional charges. Monitor the 60 requests per minute limit per key to maintain steady throughput during production.

SDK Usage ['Use the official OpenAI SDKs by pointing them to https://api.playmm33.com/v1. This works because the request shape matches the standard speech endpoint. No custom client libraries or proprietary SDKs are needed, allowing you to reuse existing infrastructure.', 'Pass your tts-... key via the Authorization header. The API accepts standard JSON bodies with model, input, voice, and response_format fields. This compatibility ensures seamless integration with your current development environment.'] ['Update base URL to api.playmm33.com/v1', 'Use standard OpenAI Python or Node SDKs', 'Include Bearer token in Authorization header']

Error Handling ['Since failed synthesis requests are free, implement retry logic for transient errors without worrying about cost overruns. The API returns standard HTTP status codes for rate limits or invalid parameters.', 'Monitor your usage dashboard to track request status. If a request fails, the charge is returned in full, allowing you to retry until success without financial penalty.']

Batch Processing ['Split long scripts into chunks of up to 4096 characters and process them sequentially. Join the resulting audio files to create complete episodes. This method handles the 60 requests per minute limit effectively.', 'For high-volume production, queue requests to stay within the rate limit. Use the speed parameter (0.25–4.0) to adjust generation time, keeping in mind that longer texts take roughly real-time duration.']

Cost Analysis

Sound wave illustration for: Cost Analysis

EchoCast bills at $10 per million characters, equating to $0.01 per 1,000 characters. Spaces and punctuation count toward this total. Failed requests incur no charge, ensuring you only pay for successful synthesis. Balances never expire, with no monthly subscriptions or minimum spend requirements.

Price Per Character

Each request processes up to 4,096 characters. Long scripts require splitting text and joining audio files. You can adjust speed between 0.25x and 4.0x to optimize character usage relative to audio duration.

Trial Usage

Signing in with Google grants $0.50 credit, covering approximately 50,000 characters valid for 7 days. Accounts created with email and password start without trial credit. This allows immediate testing of voice quality and API integration.

Top-Up Methods

Add funds exclusively via USDT on the TRON network (TRC-20). The minimum top-up is $5. Credits are automatically added within minutes of transfer. No credit cards, PayPal, or invoices are supported.

Quality Control

Sound wave illustration for: Quality Control

Quality control ensures your episodic content meets broadcast standards before distribution. Use listening tests to verify narrative flow, check for artifacts like unnatural intonation, and adjust speed settings for optimal listening comfort within the 0.25–4.0 range.

Listening Tests

Play generated audio files through standard headphones and speakers. Verify that the voice matches the intended tone, whether Aria’s warm narration or Atlas’s neutral news delivery.

Test short phrases and long scripts to ensure consistent pacing across the entire episode.

  • Test in quiet and noisy environments
  • Verify emotional tone matches script intent
  • Check end-to-end audio clarity

Artifact Detection

Listen for robotic tones or unnatural intonation, particularly at sentence boundaries. Synthetic speech carries an inaudible watermark; ensure it does not interfere with perceived audio quality.

If artifacts persist, try switching voices or adjusting the speed parameter.

  • Check for robotic tones
  • Verify no unnatural intonation
  • Ensure watermark invisibility

Speed Adjustments

Adjust the speed between 0.25 and 4.0 to optimize listening comfort. Faster speeds may introduce artifacts; slower speeds may feel draggy.

Use speed adjustments to align audio length with your podcast’s standard runtime without re-generating the entire script.

  • Adjust speed 0.25–4.0
  • Optimize listening comfort
  • Align with standard runtime

Future-Proofing

Sound wave illustration for: Future-Proofing

The API endpoint remains stable under standard load, but background processing means requests are handled sequentially. Expect slight queuing during peak usage rather than instant delivery.

Audio formats are standardized across the platform. We maintain backward compatibility for existing integrations, ensuring your scripts continue to work without immediate refactoring.

API Stability

The endpoint https://api.playmm33.com/v1 uses a consistent structure. Rate limits are predictable at 60 requests per minute, allowing reliable batch scheduling.

Input text is used only to produce the audio and is not used for training. Generated audio is returned in the response and not stored.

  • Sequential processing under load
  • Predictable 60 RPM limit

Format Updates

We support MP3, WAV, OPUS, FLAC, and PCM. These formats cover web distribution and archival needs.

No immediate breaking changes are planned for these core formats.

Voice Additions

The current eight voices are based on licensed LibriTTS-R recordings. We do not plan rapid voice additions.

New voices would require new key provisioning.

Future-Proofing

Final Recommendations

Sound wave illustration for: Final Recommendations

For news content, Atlas delivers the neutral clarity required for consistent reporting. Fiction benefits from Aria’s warm narration or Luna’s soft storytelling, ensuring emotional resonance without artificial exaggeration. These presets maintain character continuity across episodic releases, reducing the cognitive load on listeners who rely on vocal consistency to track complex plots.

Best Voice for News

Atlas is the optimal choice for news and journalism due to its neutral, authoritative tone. It avoids the energetic bias of Felix or the conversational warmth of Milo, providing a baseline sound that lets the script’s facts take center stage. This neutrality ensures the audio feels objective and professional, ideal for daily updates or serialized reporting.

  • Use Atlas for standard news readings
  • Pair with 1.0 speed for clarity
  • Avoid excessive speed adjustments to preserve tone

Best Voice for Fiction

Aria offers a warm, clear narrator voice perfect for book summaries and audiobooks. Luna provides a soft, storytelling quality ideal for immersive narratives. Both voices support long-form content efficiently, handling up to 4096 characters per request. Use Sage for instructional fiction or guides where a calm, instructional tone aids comprehension.

  • Aria for warm, clear narration
  • Luna for soft, engaging storytelling
  • Sage for instructional or guide-style fiction

Default Settings

Start with the 'tts-1-hd' model for higher fidelity, especially for long scripts. Set speed to 1.0 for natural pacing; adjust between 0.25 and 4.0 only if timing constraints require it. Use MP3 format for web distribution to balance quality and file size. Always specify the language code (e.g., 'en') for non-English text to prevent detection errors.

Ensure you split texts longer than 4096 characters to avoid request limits. For episodic series, stick to one voice per show to maintain brand identity. Use the OpenAI SDK with the base_url updated to our API endpoint for seamless integration into your existing pipeline.

  • Model: tts-1-hd
  • Speed: 1.0 (default)
  • Format: mp3
  • Language: explicit ISO code

Listen to the 8 voices

Each sample below was generated by this API from the same sentence. Stop coding and start creating your dream app with just your voice and a few simple taps.

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. Creative Commons Attribution 4.0
  3. ISO 639 language codes
  4. Node.js documentation
  5. openai-node on GitHub
  6. FFmpeg documentation
  7. W3C WAI: making audio and video accessible
  8. LibriTTS-R paper (arXiv 2305.18802)
  9. Python documentation
  10. LibriTTS-R corpus (OpenSLR 141)
  11. LibriVox public-domain audiobooks
  12. OpenAI API reference: create speech

More on this site