Basic flow
Horizon-Odia
Turn spoken Odia into searchable text for district-level service requests and radio archives. Shunya Labs Horizon Indic is designed for speech in Indian languages and regional varieties, with a platform for both uploaded recordings and live audio streams.
Odia speech to text
Odia is used in Odisha. A single clean sample rarely captures the practical demands of a deployed speech system. Real-world audio includes different microphones, spontaneous phrasing, local place names, names of people, interruptions and background noise. These details matter when a transcript feeds a support ticket, a community archive or a voice assistant.
For Odia, pay particular attention to Odia sounds, names of districts, and phone-quality audio. Select a reference script before testing, and compare the model output against transcripts checked by speakers familiar with the intended audience. A language name in a catalogue is the starting point for evaluation, not a guarantee that every dialect, script or audio channel performs identically.
Basic flow
| Capability | Horizon Indic |
|---|---|
| Odia speech recognition | |
| Horizon Indic | |
| Batch transcription | |
| Real-time streaming | |
| Automatic language detection | |
| Speaker diarization | |
| Speaker identification | |
| Word-level timestamps | |
| Confidence scores | |
| Multilingual recognition | |
| Code-Switch platform | |
| Intent detection | |
| Sentiment analysis | |
| Emotion detection | |
| Summarisation | |
| Redaction | |
| Streaming API | |
| Batch API | |
| Self-hosted deployment |
Try it yourself
Working with Odia audio requires attention to Odia sounds, names of districts, and phone-quality audio.Test your recordings, review the transcript in the writing system your audience expects,and connect the result to the workflows that matter to your team.
Horizon Indic for Odia
Supply Odia audio through a supported file upload for batch transcription, or use a streaming connection for liveexperiences. Choose the deployed Indic model and, when supported, lock the correct language route; the publiclyreferenced code is “or” (confirm the code with the production language endpoint). Receive text to review,search, display in an application, or pass to an authorised downstream workflow. For supported batch configurations,consider diarization, timestamps and optional conversation intelligence. Feature availability can depend on configuration.
Odia speech
Odia-first transcription
Odia is commonly associated with Odia script. Writing-system choices can still vary by community, context ordownstream application. Transcription captures the spoken language as text; transliteration changes the script;translation expresses the meaning in another language. These are distinct operations and should be testedseparately. Before deployment, confirm how the model represents Odia names, numbers, locally used terms andrelevant scripts. Keep a native-speaker-reviewed set of short recordings as a repeatable acceptance test.
Odia speech
Real-time Odia transcription
Horizon supports real-time speech recognition through WebSocket streaming. A live application does not need towait for the complete conversation. Partial text arrives while speech continues, then the segment is finalised.
Live audio
Microphone or application stream.
Partial and final transcripts
Streaming or batch
Use streaming when text is needed now. Use batch when the full recording already exists.The speech-recognition goal stays the same. Only the processing mode changes.
Batch
Streaming
Use real-time recognition when an application needs the user's Odia words during a conversation - for example a live caption screen, a voice-driven form or agent assistance. Shunya documents WebSocket streaming in its Horizon platform. Validate latency, audio quality and per-language recognition using the deployed service rather than assuming an aggregate benchmark applies to Odia.
Batch
Use batch transcription when you already have an Odia interview, phone recording, podcast, lecture or field recording. The system processes the completed audio and returns text that reviewers can correct and tag. This mode is a useful starting point for archives and back-office processing because reviewers can inspect the entire recording.
| Requirement | Streaming | Batch |
|---|---|---|
| Live Odia speech | ||
| Recorded Odia audio | ||
| Partial transcripts | ||
| Complete transcript | ||
| Voice agents | ||
| Post-call intelligence | ||
| WebSocket | ||
| HTTP API |
Language and model
Horizon supports automatic language detection across its multilingual speech platform. When an application alreadyknows the user is speaking Odia, set the deployed configuration accordingly. Horizon Indic lists Odia as languagecode or. Use the code the current language listing exposes for the deployment you are calling.
Automatic language detection
01
Use Horizon Indic.
02
Use the appropriate Code-Switch model where the required pair is supported.
03
Use the appropriate domain-specific speech model where it applies.
The conversation determines the recognition path
Shunya supports different Horizon model families for different speech environments.
Odia and English vocabulary
Odia speakers may use English terms inside an otherwise Odia conversation. Horizon includes Code-Switchmodels for speech that moves between languages in the same interaction. A dedicated Odia-English pairVerify Odia-English Code-Switch routing in your active deployment and test with your own mixed audio.
Safe implementation
Check the supported deployment pair, use that recognition route, and evaluate the transcript on real mixed audio before you rely on it.
Names, numbers, and references
A local organisation records an Odia service enquiry and needs a readable transcript before responding.Input: a consented Odia recording containing a real enquiry, interview or instruction.ASR: select Horizon Indic and review the transcript with an Odia speaker, checking names, dates and the expected script.Optional translation: send the reviewed transcript to Parallax only if the exact language pair has been validated.Output: a searchable record for a CRM, media archive, support dashboard, content workflow or educational resource.
Proper nouns
Operational details
Illustrative workflow only: no generated native-language sentence or translation here is presented as a measured product result.
Who said what
Speaker diarization assigns transcript segments to different speakers. Speaker identificationcan resolve registered voice profiles where it is configured. Unknown speakers stayon anonymous labels. Identification is not legal identity verification.
Diarization
SPEAKER_00 Customer request
SPEAKER_01 Agent response
SPEAKER_00 Customer follow-up
Before
SPEAKER_01 Agent response
After
Agent Priya Agent response
Timing and confidence
Verbose Horizon output can include a word start time, a word end time, and a confidence score.Confidence is an operational signal. It is not proof that a transcript word is correct.
Word-level timestamps
Confidence scores
Search
Where Odia speech becomes text
Enable an in-app listening layer so people can interact with services without typing lengthy messages. The immediate opportunity for Odia is district-level service requests and radio archives.
Connect transcripts to case management, QA review, agent assistance and downstream analytics. This is especially useful when the source material contains Odia sounds, names of districts, and phone-quality audio.
Turn spoken questions into reviewable text for community information desks and public-facing applications. Keep human review in the workflow until your acceptance thresholds have been met.
Customer support
Meetings
Odia voice agents
Phone and file audio
Odia customer conversations may arrive over ordinary phone systems. For video, the recognitionmodel follows the spoken audio rather than the video container.
Phone and telephony
Files
Phone call
Recorded file
Video
Post-call intelligence
For supported batch workflows, Horizon can add speakers, timing, intent, sentiment, emotion,a summary, and redaction. These enrichments are positioned as batch and post-callcapabilities. Sentiment supplements what the speaker said. It does not replace it.
Who spoke in each segment.
Resolve registered voice profiles where configured.
Why the conversation happened, using the organisation's own taxonomy.
A contextual signal for QA, experience analysis, escalation review, coaching, and research.
An inferred post-call signal, not a verified fact about the speaker.
Reason for contact, issue, information, action, escalation, and next step.
Connect the transcript back to the recording.
Mask configured sensitive terms. One layer of a wider data-governance strategy.
Batch today
Summarisation
Redaction
Call archives, CRM, QA, search, and analytics.
Developers
Use the Horizon ASR API for speech recognition and Parallax for text translation. For Odia, the editorial language identifier is or; obtain the accepted code and model ID from the production API. Shunya's current main documentation calls the Indic model horizon-indic, while older documentation may refer to zero-indic.
{
"file": "odia-call.wav",
"model": "horizon-indic",
"language_code": "or"
}Configure as needed
Output can include
Basic flow
Benchmarks
Broader Horizon evaluations include Hindi and other platform results. They are not an Odia word error rate, an Odia accuracypercentage, or an Odia benchmark rank. A real production sample is more meaningful than an unsupported number.
Test it on your own Odia audio before production.
Test Horizon on your Odia audioBuilt for production
A production Odia workflow may need recognition, streaming, batch processing, speaker structure,intelligence, an API, and room to add other Indian languages. Self-hosted deployment is part of thebroader production feature set where an organisation needs more control over infrastructure.Follow current Shunya deployment documentation for the exact configuration.
| Requirement | Horizon |
|---|---|
| Odia speech to text | |
| Horizon Indic | |
| Batch transcription | |
| Streaming transcription | |
| Automatic language detection | |
| Multilingual recognition | |
| Code-Switch platform | |
| Speaker diarization | |
| Speaker identification | |
| Word timestamps | |
| Confidence scores | |
| Telephony workflows | |
| Intent detection | |
| Sentiment analysis | |
| Emotion detection | |
| Summarisation | |
| Redaction | |
| API integration | |
| Multilingual expansion | |
| Self-hosted deployment |
Questions
Odia appears in the published Horizon Indic supported-language list. Test the current API with your own speakers, scripts, accents and recording conditions before production use.
Horizon offers both streaming and batch workflows. Verify the exact Odia configuration and performance in the target deployment.
Odia is associated with Odia script. Confirm that the deployed output format matches your audience and do not assume every alternate script is automatically available.
The normal route is speech transcription followed by Parallax text translation. The published evaluation reports 39 BLEU, but real-world content and direction still require evaluation.
Use consented recordings that represent your actual speakers and tasks. Review proper nouns, Odia sounds, names of districts, and phone-quality audio, and output-script consistency rather than relying only on a multilingual platform-wide benchmark.
Get started
Try real Odia audio in the Shunya Labs playground, inspect the output, and speak with the team about streaming,APIs and translation support. Bring the scripts, speaker groups, audio samples and required downstreamlanguage that your application will use.