A guide to generating natural Spanish audio with AI. Model comparisons, accent options, and tips for mixed English/Spanish content.
Spanish TTS in 2026
Spanish is the fourth most spoken language in the world, with over 500 million native speakers. For content creators, marketers, and educators producing Spanish-language material, quality text-to-speech is essential. But until recently, most TTS engines treated Spanish as an afterthought, offering a handful of robotic voices that sounded nothing like natural speech.
That has changed. Several open-source models now produce Spanish audio that sounds genuinely natural, with proper intonation, rhythm, and pronunciation. Murmur bundles three models that support Spanish: Kokoro, Qwen3 TTS, and Fish Audio S2 Pro. Each handles the language differently, and the best choice depends on your specific needs.
Which Models Support Spanish
Kokoro supports Spanish as one of its 9 languages. The output is clean, consistent, and well-paced. It handles both Latin American and Castilian pronunciation patterns depending on the voice selected. For straightforward narration (blogs, documentation, educational content), Kokoro is the reliable default.
Qwen3 TTS brings the strongest multilingual capability. Its code-switching ability is particularly valuable for Spanish content that includes English terms (common in tech, business, and marketing). A sentence that mixes "el nuevo framework de machine learning" flows naturally without the jarring accent shifts you hear from single-language models.
Fish Audio S2 Pro produces the most natural-sounding Spanish speech overall. The prosody is excellent, with pauses and emphasis that match how native speakers actually talk. The tradeoff is generation speed: Fish Audio takes 2 to 3 minutes per 1,500 words compared to Kokoro's 30 to 45 seconds.
Regional Accent Considerations
Spanish varies significantly by region. Castilian Spanish (Spain) features the distinctive "theta" sound for c/z, different vocabulary choices, and particular rhythmic patterns. Latin American Spanish encompasses its own regional variations, from Mexican Spanish to Argentine Spanish to Caribbean dialects.
In Murmur, the accent is primarily determined by the voice you select. Voices trained on Latin American Spanish data will produce Latin American pronunciation. Castilian voices will produce Castilian pronunciation. When selecting a voice, listen to a sample that includes words with c, z, and ll to quickly identify the regional accent.
For international audiences, neutral Latin American Spanish (similar to Mexican broadcast Spanish) is generally the safest choice. It is understood across all Spanish-speaking regions and avoids strongly regional features.
Comparison: Murmur vs Cloud Services for Spanish
| Feature | Murmur | Google Cloud TTS | ElevenLabs |
|---|
Privacy for Business Spanish Content
One often-overlooked advantage of local TTS: privacy for sensitive content. Business documents, legal contracts, financial reports, and internal communications in Spanish often contain confidential information. With cloud TTS services, every word is sent to external servers for processing.
Murmur processes everything locally on your Mac. For legal firms handling Spanish-language contracts, healthcare providers creating patient materials, or businesses producing internal training in Spanish, this is not just a convenience. It is a compliance requirement in many jurisdictions.
Tips for Better Spanish TTS
- Include proper accent marks (tildes, acute accents) in your text. TTS models use these to determine correct pronunciation. "ano" and "año" are very different words.
- For mixed English/Spanish content, use Qwen3 TTS. Its code-switching capability handles language transitions more naturally than other models.
- Test inverted question marks and exclamation marks (¿ ¡). Some models use these as intonation cues to adjust pitch at the beginning of a sentence.
- For numbers and dates, write them out in Spanish. "15 de abril de 2026" rather than "4/15/2026" to avoid ambiguous formatting.
- Slow the speed slightly (0.95x) for educational or formal content. Spanish has a naturally faster syllable rate than English, and slowing down improves clarity.
Frequently Asked Questions
Can I choose between Latin American and Castilian Spanish?
Yes. The accent depends on the voice you select. Murmur's voice library includes both Latin American and Castilian Spanish voices. Preview a voice with a test sentence that includes distinguishing sounds (like "cerveza" or "Barcelona") to identify the accent.
How does Murmur handle accented characters?
Murmur's models read accented characters (á, é, í, ó, ú, ñ, ü) correctly and use them for pronunciation. Always include proper diacritics in your text for the best results. Missing accents can cause mispronunciation.
Is the Spanish quality as good as the English?
English is the primary training language for most models, so English output tends to be slightly more natural. That said, Spanish quality from Fish Audio and Qwen3 is very good and suitable for professional use. The gap has narrowed significantly in 2026.
Can I mix Spanish and English in the same passage?
Yes, and Qwen3 TTS handles this best. It was specifically designed for multilingual and code-switching scenarios. Kokoro and Fish Audio also handle mixed content, though transitions between languages may be slightly less smooth.
What about other Spanish-related languages like Catalan or Galician?
Murmur does not officially support Catalan or Galician as separate languages. Some models may produce reasonable output for text in these languages due to their similarity to Spanish, but results will be inconsistent. For dedicated Catalan or Galician TTS, specialized tools are a better choice.
Try Murmur - $49 one-time. No subscriptions, no cloud, no per-character fees.
Originally published at murmurtts.com
Top comments (0)