In an age where information is constantly generated and reshaped, archiving spoken content has become critical. Whether it’s for promoting accountability, preserving institutional memory, or distinguishing fact from fiction — transcription is the backbone of transparency.
Across the globe, countless hours of audiovisual material are created daily: interviews, reports, parliamentary debates, council meetings, educational panels — all rich in voices, perspectives, and intent.
Take the example of a regional parliament: a single plenary day can generate hours of video featuring dozens of speakers. Transcribing such sessions manually, especially with domain-specific vocabulary or dialectal variation, traditionally takes up to 8 hours — per day of content. This is costly, time-consuming, and difficult to scale.
That’s where DeepVA comes in.
By combining our Deep Live Hub for real-time transcription and subtitling with the new Advanced Speech Recognition Module, spoken-word content can now be transcribed, labeled, and semantically enriched — in minutes, not hours. Even better: this includes support for Arabic and its many dialects, thanks to our new partnership with Lisan, a language technology company based in Saudi Arabia.
What does that mean in practice?
The Advanced Speech Recognition Module not only produces high-quality, speaker-labeled transcripts — it can also apply custom dictionaries for political terms, names, or ministries, and plug into Large Language Models (LLMs) for automated summaries, key quote extraction, or conversion to simple language. What once required a full workday now takes under an hour — with higher consistency and fewer errors.
This isn’t just about productivity.
It’s about impact: empowering journalists, public institutions, and communication teams to work faster, deliver more accurate content, and ultimately, strengthen democratic transparency through technology.
What’s new?
Thanks to our partnership with Lisan, a Saudi Arabian AI company, DeepVA now supports Arabic—including many of its diverse dialects. Arabic is a language with vast regional variation in vocabulary, sound and syntax. DeepVA’s system can now understand and transcribe these differences effectively.
Lisan specializes in AI-powered Arabic writing tools and proofreading. Together, we enhance Arabic-language transcription for media, education, and public institutions.
The Advanced Speech Recognition Module goes beyond basic transcription. It enables:
These features are especially valuable in contexts where clarity, accuracy, and speed are critical.
Use cases
-
Editorial Workflows & Journalism (UI or API)
-
Parliamentary and Administrative Documentation (API)
Newsrooms can streamline their editorial process by using the Advanced Speech Recognition module in DeepVA to quickly convert interviews, press conferences or live broadcasts into searchable text. With speaker labels, quotes can be correctly attributed, eliminating the need for manual transcription and reducing errors.
Here’s how it works (UI):
- Upload an interview or report.
- Use the Advanced Speech Recognition with your custom dictionary.
- Lecture & Export transcript directly from the interface into a word document, with timecodes or without
How it works (API):
- Integrate transcription directly into your Newsroom or CMS system.
- Push a video or audio recording to our API.
- The asset gets transcribed with our Advanced Speech Recognition and your custom dictionary.
- Fetch finished results without having the users leave your UI and directly let them work with the text.
Public institutions often deal with large-scale, recurring recordings, such as parliamentary debates, council meetings, and public hearings. When citizens cannot easily verify what their elected representatives said—or did not say—during a plenary session, transparency suffers, and confidence might erode. DeepVA’s Advanced Speech Recognition Module bridges the gap between raw audio and an auditable public record.
In the future, when using the Deep Live Hub for subtitling, you will be able to forward the transcript to our Transcript Editor. This integration enables institutions to automate not only their subtitling for accessibility, but also their entire documentation workflow, ensuring consistent output and traceability.
How it works (API):
- When a new recording is recorded, an API call sends the finished file for additional analysis.
- Custom dictionaries include party names, speaker lists, or legal references, and additional metadata are applied.
- Transcripts are returned as structured data (e.g., JSON, XML, DOCX) and auto attached to the session archive or published.
