speech-to-text Arkiv

In Live Broadcast TV, speech-to-text technology changes fast. A model that performs best today may not be the best option tomorrow, or for the next language you need to support.

That is one of the core ideas behind Baruch.

Why We Built Baruch to Be ASR-Agnostic

Baruch is built with an ASR-agnostic architecture. Instead of locking broadcasters into a single speech recognition provider, the software can integrate with different ASR engines depending on language, workflow, security requirements, or editorial needs.

ASR-Agnostic Live Subtitling with Multiple AI Speech-to-Text Engines

Some AI models perform better in Swedish, others in English, Arabic, or multilingual broadcasts. Today, providers like Whisper or Speechmatics may lead in certain scenarios, but the landscape is evolving constantly. Broadcasters should not have to rebuild their subtitle workflows every time the market changes.

Flexible Broadcast ASR Subtitling Workflows

By separating the live captioning environment from the speech recognition engine itself, Baruch gives broadcasters flexibility to choose the right AI for each production and to adapt over time without replacing the entire system.

This approach also opens the door to more advanced multilingual workflows. We are currently developing translation features that allow live speech to be automatically translated into additional languages in real time. A Swedish programme could instantly generate English subtitles, or multilingual discussions could be translated into Swedish for domestic audiences.

Live subtitles for linear TV broadcasting services.

Television screen displaying live subtitles with closed captioning for TV viewers.

Real-Time Translation and Multilingual Live Subtitles for the best available systems

The idea is simple: connect the best available technologies together instead of depending on a single vendor stack.

In broadcast environments, flexibility is not just a technical preference. It is long-term operational resilience.

More about Baruch and live subtitles for linear TV:
Technical overview
Live subtitles for linear TV

Contact

Txtplay logo blue

Contact sales for a demo

Txtplay Automatic Speech Recognition AI Enhancements

Txtplay handles AI tools such as ASR (Automatic Speech Recognition), also known as speech-to-text. Txtplay will always use the ASR version with the highest accuracy and quality. Txtplay can also integrate with our partner’s video platforms, allowing our customers to utilize the best speech-to-text AI directly within their video platform for a streamlined workflow.

Benefits of Txtplay Automatic Speech Recognition AI for Video

With automatic speech recognition (ASR), Artificial Intelligence (AI) recognizes the words spoken in your video and delivers machine-based texts for indexing, subtitling, and searching. The result is good and usable, but not always perfect.

Since Friday the 20th of October, you’ll notice remarkable enhancements with Txtplay for the following languages:

Cantonese (yue), Catalan (ca), Czech (cs), Finnish (fi), Greek (el), Hindi (hi), Hungarian (hu), Malay (ms), Polish (pl), Russian (ru), Thai (th), Vietnamese (vi), Arabic (ar), Danish (da), Dutch (nl), Italian (it), Japanese (ja), Korean (ko), Mandarin (cmn), Norwegian (no), Portuguese (pt), Swedish (sv)

What exactly will you notice on the new recognition?

Enhanced Accuracy: Get ready for even more accurate transcriptions!
Speaker Diarization: Improved speaker distinction in multi-speaker content.
Faster Transcriptions: Lightning-fast results for a more efficient workflow.

Please tell us your opinion

Tag Archive for: speech-to-text

Baruch: An ASR-Agnostic Approach to Live Subtitling for Broadcast TV