--- title: "Features" slug: "sr-features" description: "Explore Knovvu SR's robust speech recognition features including real-time transcription, accent coverage, custom vocabulary, and multilingual model support." updated: 2026-04-16T04:47:07Z published: 2026-04-16T04:47:07Z --- > ## Documentation Index > Fetch the complete documentation index at: https://docs.knovvu.com/llms.txt > Use this file to discover all available pages before exploring further. # Features | Document Number | Revision Number | Revision Date | | --- | --- | --- | | KN. GU.25.EN | Rev34 |13.04.2026 | ### Language Coverage * **Support for 99+ Languages:** More than 99 languages are available, enabling organizations to build speech-enabled experiences for a wide range of geographies, markets, and customer segments. * **Multilingual and Bilingual Models:** Multilingual and bilingual model options leverage knowledge from multiple languages, helping improve performance in multilingual and mixed-language scenarios. * **Accent Coverage in a Single Model:** Multiple accents of the same language can be handled within one model, reducing operational complexity and eliminating the need to manage separate accent-specific models. --- ### Model Flexibility and Technology Foundation * **Support for Hosting Different Model Foundations:** Different speech recognition model foundations, including approaches such as Whisper and Dolphin based models, can be hosted to match language, use case, and performance expectations more effectively. * **Flexible Architecture for Evolving Speech Technologies:** A flexible technology foundation makes it possible to adapt recognition solutions as speech technologies evolve, rather than being limited to a single fixed model strategy. --- ### Vocabulary and Model Adaptation * **Custom Word Support:** Domain-specific, business-specific, or customer-specific words that are not currently included in the language model can be added upon request. * **Fine-Tuning with Real Customer Data:** Models can be improved through fine-tuning with real customer data, allowing better adaptation to customer terminology, speaking habits, domain language, and real-life acoustic conditions. * **Domain Adaptation for Industry Needs:** Sector-specific terminology and enterprise scenarios can be addressed through tailored adaptation, helping improve recognition quality in areas such as contact centers, banking, telecom, and public services. --- ### Language Model Development * **Model Creation for Low-Resource Languages:** Dedicated speech recognition models can be developed even for languages with little or no ready training data by working with customer-provided or specially collected datasets. * **Custom Data-Based Model Training:** For languages without an existing mature model, custom development can be carried out with sufficient language data, typically requiring large-scale datasets such as 200+ hours depending on the language and target quality. * **High-Potential Accuracy for New Language Models:** With adequate, high-quality training data, custom speech recognition models can achieve strong performance levels, including 85%+ success rates for previously unsupported or low-resource languages. --- ### Audio and Input Support * **Wide Audio Format Compatibility:** Audio conversion is handled through ffmpeg, with support for major formats such as G729, MP3, MP4, WAV, and Opus. * **Flexible Audio Input Handling:** Different audio sources and integration flows can be accommodated, making it easier to connect speech recognition into existing telephony and application environments. --- ### Text Processing and Output Control * **Numeral & Entity Formatting:** Recognized content can be transformed into more usable written output by converting spoken numerals and selected entities into normalized text representations, improving readability and downstream system usability. * **Masking of Sensitive Information:** Sensitive information can be masked in recognition output through user-defined regex rules, helping protect transcribed content. * **Structured Recognition Output:** Transcribed speech can be delivered in a format suitable for automation, analytics, reporting, and operational workflows. * **Time-Aligned Transcription:** Transcription output can be provided together with timing information, supporting use cases such as subtitle generation, audio-text synchronization, search within recordings, analytics, and detailed post-processing. --- ### Integration and Deployment * **Easy Integration with APIs and SDKs:** User-friendly APIs and SDKs help simplify integration into existing applications and platforms, reducing implementation effort. * **Flexible Integration Options:** Different integration methods and communication structures can be used to align with varying architectural and operational needs. * **Cloud and On-Prem Deployment:** Deployment can be made in cloud or on-prem environments, depending on infrastructure, security, and compliance requirements.