Central Mnong Speech to Text: Preserving a Language Through AI

Central Mnong (ម្នង) is an Austroasiatic language spoken by around 100,000 people in the highlands of Cambodia and Vietnam. As a member of the Bahnaric branch, it is closely related to languages like Bunong and Stieng. Despite its rich oral tradition, Central Mnong has a very limited written presence — most speakers use Khmer or Vietnamese for literacy. The Khmer script is increasingly used to write Mnong in Cambodia, but there is no standard orthography, and most written materials are produced by missionaries or linguists.

Why Accurate Speech-to-Text Matters for Mnong

Without a robust transcription tool, Mnong speakers are often excluded from the digital world. Voice notes, community videos, and oral histories remain trapped in audio form. Accurate speech-to-text can:

• Preserve endangered oral traditions for future generations.
• Enable Mnong speakers to create subtitles for their content on YouTube and Facebook.
• Assist linguists in documenting the language's unique phonology.
• Provide accessibility for deaf Mnong speakers via captions.

The Challenge of Transcribing Central Mnong

Central Mnong presents several unique challenges for ASR:

• Phonation contrast: The language distinguishes words by breathy vs. clear voice. For example, /maː/ (clear) vs. /ma̤ː/ (breathy) can change meaning. Most ASR systems cannot detect this.
• Adapted Khmer script: Khmer does not have dedicated letters for Mnong's preglottalized stops, so transcription must use a consistent mapping.
• Dialect diversity: Four major dialects (Biat, Preh, Bu Nong, Rlom) differ significantly. A model trained on one dialect may fail on another.
• Code-switching: Many Mnong speakers mix in Khmer words, especially for modern concepts like "phone" or "school".

Use Cases in Practice

• Podcasters: A Mnong-language podcast about traditional medicine can now be automatically transcribed and shared as a blog post with Khmer script text.
• Researchers: A linguist recording elder speakers can upload the audio and get a draft transcript in minutes, saving weeks of manual transcription.
• Activists: Community leaders can add subtitles to videos documenting land rights issues, reaching a wider audience.

How Speechyou Helps

Speechyou is the first speech-to-text tool to support Central Mnong with Khmer script output. Our model is trained on a corpus of spoken Mnong from multiple dialects, with special attention to breathy phonation. We use a custom orthographic mapping that follows the conventions used by Mnong writers in Cambodia. The system is available as a web app and can be used for free (with unlimited transcription on the Solo plan).

Key Features

• Real-time transcription: Upload audio or video and get text in minutes.
• Subtitle export: Generate SRT and VTT files for video platforms.
• Dialect adaptation: Choose your dialect from a list or let the model auto-detect.
• Continuous improvement: User corrections help the model learn.

Getting Started

Transcribing Central Mnong with Speechyou is simple. Just upload an audio file (MP3, WAV, or video) and select "Central Mnong (Khmer script)" as the language. After a few seconds, you'll receive a transcript with timestamps. You can edit the text and export subtitles or a plain text file. The tool works on desktop and mobile browsers, so you can even record directly from your phone.

Conclusion

Central Mnong is a vital part of Cambodia's cultural heritage, but it faces the threat of language shift. By providing accurate speech-to-text and transcription services, Speechyou empowers Mnong communities to document, share, and preserve their language in the digital age. Whether you are a community elder, a content creator, or a researcher, our tool makes it easy to turn spoken Mnong into written words.