Mapudungun Speech to Text: Preserving a Living Language with AI

Mapudungun, the language of the Mapuche people, is spoken by around 260,000 individuals in Chile and Argentina. It is a language isolate with no known relatives, making it a unique linguistic treasure. Despite its relatively small speaker population, Mapudungun has a rich oral literature, including epic poems like the Ngillatun and countless stories passed down through generations. Accurate speech to text for Mapudungun is not just a technical challenge — it is a tool for cultural preservation and revitalization.

Why Accurate Transcription Matters

For Mapudungun speakers, transcription opens doors to digital media, education, and legal recognition. Podcasts, radio programs, and YouTube videos in Mapudungun can reach wider audiences when subtitled. Researchers studying the language's complex morphology need precise transcripts to analyze verb forms and discourse patterns. Community organizations working on language revitalization can create teaching materials from recorded conversations. Without reliable speech to text, these efforts rely on manual transcription, which is slow and costly.

Transcription Challenges Specific to Mapudungun

Mapudungun presents several hurdles for automatic speech recognition:

• Polysynthetic morphology: Verbs can contain up to 15 morphemes. For example, pewmalen (to dream about oneself) combines the root pewma with reflexive and infinitive suffixes. The ASR must handle long word forms without splitting them incorrectly.
• Ejective consonants: Sounds like /p'/, /t'/, and /k'/ are produced with a glottalic egressive airstream, which is rare globally and can be mistaken for regular stops by generic models.
• Dialectal variation: The four main dialects differ in pronunciation. The Huilliche dialect, for instance, tends to soften consonants, while Moluche retains more conservative forms.
• Limited digital resources: Few publicly available audio datasets exist, requiring specialized training techniques.

Speechyou addresses these challenges with a dedicated Mapudungun model that uses transfer learning from related polysynthetic languages and a custom tokenizer for agglutinative words. The model is trained on a balanced dialect corpus and fine-tuned to recognize ejective consonants and other distinctive features.

Use Cases for Mapudungun Speech to Text

• Oral history preservation: Transcribe interviews with elders to document traditional knowledge, myths, and historical accounts.
• Media subtitling: Generate SRT subtitles for Mapudungun-language videos on YouTube or community TV.
• Language education: Create transcripts of classroom lessons and traditional stories for learners.
• Academic research: Analyze spontaneous speech for studies on syntax, discourse, and language change.
• Accessibility: Provide text alternatives for deaf or hard-of-hearing Mapudungun speakers.
• Community radio: Automatically transcribe radio broadcasts for archiving and distribution.

How Speechyou Helps

Speechyou offers a seamless solution for Mapudungun transcription. Upload audio or video files, and the AI returns accurate text within minutes. The platform supports export to SRT, VTT, and plain text formats, making it easy to integrate with video editing tools. With unlimited transcription included in the Solo plan, there is no per-minute cost — ideal for long oral history recordings or large media projects. Whether you are a linguist, educator, or community activist, Speechyou empowers you to transcribe Mapudungun audio with confidence.

The Future of Mapudungun in the Digital Age

As more Mapuche communities embrace digital media, the demand for Mapudungun speech to text will grow. Speechyou is committed to supporting this language with ongoing model improvements and dialect coverage. By making transcription accessible, we help ensure that Mapudungun remains a vibrant part of the world's linguistic heritage.