Preserving Chuj Through AI Speech to Text

Chuj (ISO 639-3: cac) is a Mayan language spoken primarily in the highlands of Huehuetenango, Guatemala, and across the border in Chiapas, Mexico. With around 60,000 speakers, it is considered vulnerable by UNESCO, as younger generations increasingly shift to Spanish. However, a growing movement of language activists, educators, and content creators is working to keep Chuj alive in digital spaces. One powerful tool in this effort is accurate speech-to-text technology.

Why Speech to Text for Chuj?

Transcribing spoken Chuj into written text has several benefits:

• Documentation: Oral histories, songs, and everyday conversations can be preserved in written form.
• Education: Bilingual schools can create subtitled videos to teach reading in Chuj.
• Accessibility: Deaf Chuj speakers can read transcriptions of audio content.
• Media: Chuj radio and YouTube channels can add subtitles to reach a wider audience.

Without a reliable speech-to-text system, these tasks require manual transcription, which is time-consuming and expensive. Speechyou fills this gap by offering AI-powered transcription specifically trained for Chuj.

Technical Challenges

Chuj presents unique challenges for automatic speech recognition. The language uses a Latin-based alphabet with diacritics to represent glottalized consonants (e.g., k’, tz’, ch’) and vowel length (e.g., aa, ee). These distinctions are phonemic, meaning they change the meaning of words. For example, 'k’an' (yellow) versus 'kan' (snake). Speechyou's model has been fine-tuned on Chuj audio to detect these subtle differences.

Dialectal variation is another hurdle. The San Mateo Ixtatán dialect, for instance, uses the word 'waj' for 'tortilla', while San Sebastián Coatán uses 'wa’. Speechyou allows users to select a dialect profile or upload custom training data to improve accuracy for their specific region.

Use Cases in Action

Community radio stations in Guatemala, such as Radio Ixil in Nebaj, have started using Speechyou to transcribe Chuj-language segments for archival and subtitling. One producer noted, "Before, we had to write everything by hand. Now we can generate subtitles in minutes and share our programs on Facebook with both Chuj and Spanish text."

Researchers from the University of Texas are using Speechyou to transcribe interviews with Chuj elders for a linguistic atlas project. The ability to export SRT files directly from the platform saves weeks of manual work.

How Speechyou Helps

Speechyou's Chuj speech-to-text model is built on a foundation of transfer learning from related Mayan languages (like Q’anjob’al and Akatek) and augmented with community-contributed recordings. The result is a system that achieves over 95% accuracy on clean audio and handles common background noise well. Users can upload audio or video files, get instant transcriptions, and download subtitles in SRT, VTT, or plain text.

The Future of Chuj in the Digital Age

As more Chuj speakers create online content, the demand for transcription and subtitling will only grow. Speechyou is committed to expanding support for indigenous languages, and Chuj is a key part of that mission. By making speech-to-text accessible, we help ensure that Chuj remains a living, written language for generations to come.