Yucuna Speech to Text: Preserving an Amazonian Language with AI

Yucuna is an Arawakan language spoken by approximately 1,000 people in the Colombian Amazon, primarily along the Caquetá and Mirití-Paraná rivers. It is a language rich in oral tradition, from mythic narratives passed down through generations to the intricate songs of shamans. Yet, like many indigenous languages, Yucuna faces pressure from Spanish and a lack of written materials. Automatic transcription tools can help document, teach, and share Yucuna in ways that paper never could.

Why accurate speech-to-text matters for Yucuna

For a language with few speakers, every tool that strengthens its presence in the digital world is a victory. Yucuna speech to text allows:

• Elders to record stories without needing to write them down themselves.
• Bilingual schools to produce reading exercises in the standard orthography.
• Researchers to quickly analyze phonological patterns and grammatical structures.
• Community members to add Yucuna subtitles to videos that reach a broader audience.

Speechyou’s Yucuna transcription engine was built from the ground up to handle this language’s specific challenges.

Specific transcription challenges in Yucuna

Yucuna phonology presents several obstacles for automatic speech recognition:

• Vowel length phonemicity: /i/ and /iː/ (spelled 'i' and 'ii') distinguish words like pii (you) and pi (a type of tree). Missing length changes meaning.
• Prenasalized stops: Sounds like /mb/ in mba (what) and /nd/ in ndu (to go) must be kept distinct from simple /b/ and /d/.
• Pitch accent: Stress is contrastive in some contexts (e.g., kái (fire) vs. kaí (monkey?)), though not fully tonal.

Speechyou’s model was trained with minimal pairs and pitch accents annotated by native speakers. It handles code-switching with Spanish gracefully, outputting Yucuna portions in the standard Latin-based orthography that marks long vowels with doubled letters.

Real-world use cases

Oral history preservation

Organizations like the Instituto Colombiano de Antropología e Historia (ICANH) can use Yucuna transcribe to digitize audio archives of elder narratives. Subtitles in Yucuna and Spanish make these resources accessible to younger generations and to the global academic community.

Education and literacy

Bilingual teachers in La Pedrera and nearby settlements can upload classroom recordings to instantly get Yucuna text. They then create worksheets and storybooks, reinforcing reading and writing skills in the native language. The auto-generated SRT files allow subtitling of educational videos on platforms like YouTube.

Community media

Local radio stations often broadcast in Yucuna. With Speechyou, they can publish transcripts of interviews and announcements on their websites, reaching listeners who prefer reading. Subtitled videos of festival celebrations and environmental workshops gain wider viewership.

How Speechyou makes a difference

Most commercial speech-to-text services ignore Yucuna completely. Google Speech-to-Text, Amazon Transcribe, and Rev AI list hundreds of languages but not a single Arawakan one. Whisper (OpenAI) covers 99 languages but Yucuna is absent. Speechyou fills this gap with a dedicated, continuously improving model. The Yucuna ASR is available in the Solo plan at no extra per-minute cost, making it accessible for non-profits and community initiatives.

The future of Yucuna in the digital age

With real-time Yucuna speech to text, the language can thrive beyond oral tradition. Young Yucuna speakers can create content on their own terms — video essays, music clips, or social media posts — and automatically add their language’s text. Each transcription reinforces the orthography and normalizes the written form in daily life. Speechyou is proud to support the Yucuna community and to help ensure that this Amazonian language remains spoken, written, and heard for generations to come.