One API
Spanish today. Another global language tomorrow.
The application does not need a completely different ASR architecture for each language.
Horizon · Spanish
Convert spoken Spanish into accurate, structured text for voice agents, customer conversations, meetings, media, contact centers, captions, analytics, and real-time AI applications. Use Shunya Horizon Universal to transcribe Spanish speech with streaming and batch recognition, speaker labels, timestamps, multilingual understanding, and optional conversation intelligence.
Hola, llamo porque hice el pago ayer, pero todavía no he recibido la confirmación.
Speaker diarization · Horizon UniversalSpanish STT
Spanish is spoken across many countries, accents, industries, and conversational environments. Real speech may contain regional pronunciation, fast conversation, English terminology, names and addresses, phone-line audio, background noise, multiple speakers, interruptions, and technical vocabulary. Horizon Universal includes Spanish in Shunya's global speech-recognition coverage.
204+
Languages supported across the current Horizon platform
3.10%
Composite WER across Shunya's 8-benchmark OpenASR evaluation
Sub-500ms
Streaming transcription across the current product platform
#1
OpenASR composite positioning published by Shunya
| Capability | Horizon · Spanish Speech to Text |
|---|---|
| Spanish Language Recognition | |
| Batch Transcription | |
| Real-time Streaming | |
| Automatic Language Detection | |
| Speaker Diarization | |
| Speaker Identification | |
| Word-level Timestamps | |
| Confidence Scores | |
| Multilingual Recognition | |
| Code-switching | |
| Intent & Sentiment | |
| Summarisation | |
| Private Deployment Options |
Spanish model: Use Horizon Universal for Spanish and other global languages.
Try it yourself
A clean prepared recording cannot tell you how ASR will perform on your customer calls,regional accents, names, addresses, product terminology, or noisy conversations.
Global speech recognition
Horizon Universal is Shunya's model for broad multilingual speech recognition, spanning 204+ languages and dialects worldwide. Spanish appears explicitly inShunya's international-language documentation alongside languages such as:
One API
The application does not need a completely different ASR architecture for each language.
Benchmarks
Shunya publishes 3.10% composite WER across 8 OpenASR benchmarks. That evaluation includes datasets such as:
3.10%
Composite WER across 8 OpenASR benchmarks
Important: These benchmarks show Horizon's overall performance. They are not a Spanish-only benchmark: Shunya has not yet published a Spanish-specific word error rate.
Spanish support: Spanish is supported by Horizon Universal. The figures above are composite results across benchmarks, not a Spanish score.
The right test: Upload the Spanish conversations your application actually receives.
Test Horizon on your Spanish audioSpanish accents
Spanish is spoken across many regions. Pronunciation, rhythm, vocabulary, and conversational style can vary across:
Real-world variation
One speaker may say computadora while another may prefer ordenador. A customer may speak slowly and formally. Another may speak quickly and conversationally.
ASR needs to handle
Global model
Horizon Universal is the appropriate Shunya model path for Spanish speech.
Real-world Spanish
Production Spanish audio can include pauses, restarts, fillers, fast speech, critical names and numbers, and noisy environments.
Quería saber... bueno, si mi pedido ya salió.
Necesito cambiar... en realidad, quiero cancelar mi pedido.
Bueno... Pues... Eh...
Speakers may join words together during natural conversation.
Customer names, addresses, cities, and brands may be the most important words in the transcript.
Phone numbers, order IDs, amounts, dates, and times.
Contact centers, stores, traffic, homes, and public places.
Speech recognition needs to work on conversation, not just carefully read Spanish.
Spanish + English
Spanish-speaking users often include English terminology, especially in:
Mi payment ya fue procesado pero todavía no recibí el confirmation email.
La app abre, pero el login button no funciona.
Quiero cambiar mi subscription al Premium plan.
One conversation
Shunya's Horizon Code-Switch model is built for multilingual speech without requiring separate language detection and model handoffs.
Important: Shunya publicly lists many multilingual combinations, but not every possible pair is available individually. For production Spanish-English code-switching, test your real audio and confirm the appropriate Horizon Code-Switch routing available in your deployment.
Real-time Spanish transcription
Some applications need the transcript immediately. Horizon supports streaming speech recognition through WebSocket connections.
Live audio flow
Text appears while speech continues. Segments are finalised as the conversation progresses. Use batch transcription when the complete recording is already available.
Batch vs streaming
Streaming
Batch
| Requirement | Streaming | Batch |
|---|---|---|
| Live Spanish speech | ||
| Recorded audio | ||
| Partial transcripts | ||
| Complete transcript | ||
| Voice agents | ||
| Post-processing | ||
| WebSocket | ||
| HTTP API |
Who said what
A Spanish customer call may look like this: SPEAKER_00 Hola, quiero revisar el estado de mi pedido.SPEAKER_01 Claro. ¿Puede darme el número del pedido? SPEAKER_00 Sí, es 48217.
Without diarization
Hola, quiero revisar el estado de mi pedido. Claro. ¿Puede darme el número del pedido? Sí, es 48217.
With diarization
SPEAKER_00 Hola, quiero revisar el estado de mi pedido.
SPEAKER_01 Claro. ¿Puede darme el número del pedido?
SPEAKER_00 Sí, es 48217.
Useful for
Customer calls
Meetings
Interviews
Podcasts
Research
Sales conversations
Support analytics
Speaker identification
Diarization answers:
Who spoke when?
Speaker identification can resolve known registered voices where configured.
Before
Voy a revisar su cuenta.
After
Voy a revisar su cuenta.
Workflow
Unknown speakers can remain anonymous labels.
Word-level timestamps
Verbose Horizon output can include start time, end time, and confidence score per word. Use timestamps for captions, search, call review, meeting playback, media editing, transcript highlighting, subtitles, and training data.
| Word | Start | End | Confidence |
|---|---|---|---|
| gracias | 1.14s | 1.48s | 0.98 |
| por | 1.49s | 1.60s | 0.99 |
| llamar | 1.61s | 1.98s | 0.97 |
Search workflow: "cancelar pedido" → find transcript match → jump directly to the moment in the audio.
From Spanish transcript to intelligence
For supported batch workflows, Shunya's intelligence layer can enrich a Spanish transcript.
Who spoke when?
Which registered speaker was talking?
What does the speaker want?
Is the interaction positive, neutral, or negative?
What dominant emotion appears in each segment?
What happened in the conversation?
Normalise important domain vocabulary.
Connect text back to the audio.
Mask configured sensitive terms.
Mask profane transcript content.
Enrichment pipeline
Spanish voice agents
A Spanish voice agent first needs to understand what the user says. Horizon becomes the listening layer of the Spanish conversational AI stack.
STT → LLM → TTS
Real-world use cases
Convert Spanish customer speech into machine-readable text in real time.
Transcribe Spanish calls for quality, analytics, summaries, search, and agent workflows.
Create searchable Spanish transcripts with speaker separation and timing.
Transcribe Spanish podcasts, interviews, webinars, videos, and broadcasts.
Capture spoken Spanish interactions for multilingual customer-service workflows.
Understand Spanish customer questions in stores, kiosks, and voice interfaces.
Transcribe Spanish customer-service conversations and account-related interactions.
Convert Spanish lessons, discussions, and lectures into searchable text.
Transcribe Spanish interviews, surveys, and qualitative research sessions.
Provide text output for spoken Spanish content.
Audio & video input
Horizon batch transcription accepts common audio formats. Shunya's batch workflow also supports common video formats by extracting the audio track first.
Video workflow
Spanish transcription works across calls, interviews, meetings, media files, and video content.
Developers
Use Horizon Universal for Spanish. Model: horizon-universal. Language: Spanish (es). You can also leave language detection on automatic when the language is not known in advance. When you already know the audio is Spanish, explicitly locking the language can reduce ambiguity, especially for short clips. Use GET /languages for the current supported language names, ISO codes, and scripts.
{
"model": "horizon-universal",
"language_code": "es",
"file": "spanish-call.wav"
}Input
Audio file, URL, or live stream.
Model
Language code es.
Configure
Output
Example logic
Choose the right model
Spanish-first speech
Use for ordinary Spanish transcription and broader global-language audio.
Best for:
Mixed-language speech
Use where conversations naturally switch languages and the relevant combination is supported by the deployment.
Best for:
Clinical speech
Use for healthcare-oriented transcription when the conversation requires the medical model and its supported language workflow.
Best for:
Model follows the audio
| Audio | Model |
|---|---|
| Spanish | Horizon Universal |
| Mixed multilingual speech | Horizon Code-Switch |
| Specialised clinical use case | Appropriate domain model |
Built for production
A clean Spanish sample can make almost any ASR system look useful. Production audio is harder.Teams need streaming, multilingual coverage, speakers, timestamps, APIs, intelligence,deployment options, and the ability to handle the audio users actually produce.
| Requirement | Horizon |
|---|---|
| Spanish Speech to Text | |
| Horizon Universal | |
| Spanish Language Code (es) | |
| Batch Transcription | |
| Streaming Transcription | |
| Automatic Detection | |
| Speaker Diarization | |
| Speaker Identification | |
| Word-level Timestamps | |
| Confidence Scores | |
| Multilingual Recognition | |
| Code-switching Platform | |
| Intent Detection | |
| Sentiment Analysis | |
| Emotion Detection | |
| Summarisation | |
| Redaction | |
| API Integration | |
| Enterprise Deployment |
Production output
Questions
Spanish Speech to Text uses automatic speech recognition to convert spoken Spanish into written text.
Yes. Spanish is explicitly included in Shunya's international-language speech-recognition coverage.
Use Horizon Universal for Spanish and other global languages.
The standard ISO 639-1 language code is es. Shunya's API supports language hints and also provides GET /languages for the current routable language list.
Yes. Horizon supports automatic language detection.
Horizon Universal spans 204+ languages and dialects. For exact deployable coverage, use the live GET /languages endpoint or the language list exposed by your deployment.
Not yet. Shunya publishes a 3.10% composite WER across 8 OpenASR benchmarks; it is not a Spanish-only figure.
Yes. Horizon supports WebSocket streaming transcription.
Yes. Speaker diarization can separate participants into speaker-labelled transcript segments.
Yes. Registered voice profiles can be combined with speaker identification where configured.
Yes. Verbose transcript output can contain word-level start time, end time, and confidence information.
Shunya's batch intelligence layer supports summarisation and other transcript enrichment. Test the feature on your Spanish production content to confirm the quality needed for your workflow.
Shunya's intelligence layer includes intent, sentiment, emotion, summary, and other optional enrichment features.
Yes. Real Spanish frequently contains English terminology. For heavily code-switched conversations, use the Horizon Code-Switch model where the relevant combination is supported and test your own audio.
Yes. Shunya's batch transcription workflow accepts supported video containers, extracts the audio track, and transcribes the speech.
Yes. Horizon provides HTTP batch transcription and WebSocket streaming.
Get started
Don't judge Spanish ASR from our prepared recording. Test yours. Try a customer call, regional accent,Spanish-English conversation, interview, meeting, podcast, video, voice note, or noisy real-worldrecording. See the transcript, speakers, timestamps, and structured output for yourself.