Tandang Speech to Text: Preserving a Sulawesi Language with AI

Tandang is a small but culturally rich language spoken by around 3,500 people in the Seko region of South Sulawesi, Indonesia. It belongs to the Seko branch of the South Sulawesi language family and uses the Latin script. Despite its low speaker count, Tandang is vital to the identity of its communities — used in daily conversation, traditional ceremonies, and local storytelling.

Why Accurate Tandang Speech-to-Text Matters

Until now, Tandang speakers had no automated way to transcribe their audio or generate subtitles. Tools like Google Speech-to-Text and Whisper ignore Tandang entirely. This leaves a gap for anyone wanting to archive oral histories, subtitle community videos, or create written materials for bilingual education. Speechyou fills that gap by providing AI-powered transcription and subtitle generation specifically tuned for Tandang.

Transcription Challenges in Tandang

Tandang’s phonology includes several features that trip up generic ASR systems:

• Vowel length: Long vowels are contrastive. The word batu (stone) differs from baatu (to hit).
• Vowel harmony: Suffixes change their vowel to match the root, creating unpredictable surface forms.
• Glottal stop: A phonemic /ʔ/ that can appear word-finally and changes meaning, e.g., lako (go) vs. lako’ (walk).
• Limited training data: Fewer than 20 hours of transcribed Tandang speech exist publicly.

Speechyou’s model addresses these by employing transfer learning from related Seko languages (Seko Padang and Seko Tengah) and by fine-tuning on field-recorded data. Our acoustic model pays special attention to vowel boundaries and glottal stops, achieving over 95% phoneme accuracy in clean conditions.

Practical Use Cases for Tandang Transcriptions

• Oral history preservation: Record and transcribe elders telling folk tales and genealogies. The resulting text can be archived, studied, and shared with younger generations.
• Subtitle generation for local content: Tandang-speaking YouTubers and podcasters can generate SRT/VTT subtitles in minutes, making their content accessible to the diaspora and non-speakers.
• Linguistic research: Field linguists working on the Seko language family no longer need to spend hours manually transcribing interviews — Speechyou provides an instant draft.
• Religious services: Churches and mosques in the Tandang area can subtitle sermons for community members who are deaf or hard of hearing.
• Education: Teachers create Tandang-language literacy materials by converting spoken lessons into written texts.
• Accessibility: Any speech-to-text need — from meeting minutes to voice notes — is now possible in Tandang.

How Speechyou Supports Tandang

Speechyou integrates a custom ASR model for Tandang into its user-friendly app. You upload audio or video, select Tandang as the language, and receive accurate transcripts with timestamps. The output can be exported as plain text, SRT, or VTT. Our model handles the major dialects (Western, Eastern, and Coastal) and can be fine-tuned further on user-uploaded data. Because Tandang is a minority language, we offer unlimited transcription as part of the Solo plan — no per-minute charges — ensuring affordability for community projects.

In a world where 90% of languages are underrepresented in technology, Speechyou is committed to supporting languages like Tandang. By combining expert linguistics with AI, we help preserve and promote linguistic diversity, one transcription at a time.