Kayapó Speech-to-Text: Preserving the Amazon’s Mebêngôkre Language with AI

The Kayapó people, known in their own language as Mebêngôkre, inhabit a vast territory in the Brazilian Amazon spanning the states of Pará and Mato Grosso. Their language, Kayapó (ISO 639-3: txu), is a member of the Macro-Jê family and is spoken by an estimated 8,000 to 12,000 individuals. Despite its relatively small speaker base, Kayapó carries immense cultural weight: it is the vehicle for oral histories, ceremonial songs, and environmental knowledge passed down through generations. However, like many Indigenous languages, it faces pressure from Portuguese and lack of digital infrastructure. Accurate Kayapó speech-to-text tools can help reverse language decline by making transcription, subtitling, and archiving accessible to communities and researchers.

Why speech-to-text matters for Kayapó

Manual transcription of long-recorded narratives is time-consuming and requires trained linguists. With AI-driven Kayapó speech-to-text, community members can instantly convert audio from meetings, interviews, and cultural events into searchable text. This supports:

• Language documentation: Elders’ stories become written records for future generations
• Bilingual education: Teachers produce classroom materials with accurate Kayapó orthography
• Media production: Add Kayapó subtitles to documentaries or YouTube videos
• Legal advocacy: Transcribe land rights discussions with government agencies

The transcription challenges unique to Kayapó

Kayapó phonology presents distinct obstacles for generic ASR systems. The language has a rich system of nasal vowels (ã, ẽ, ĩ, õ, ũ, ə̃) that are phonemic; confusing kãt (to see) with kôt (to go) changes the meaning entirely. Many commercial speech engines fail to distinguish these because they treat nasalisation as a vowel quality rather than a phonemic feature. Speechyou’s acoustic model includes a dedicated classifier for vowel nasality, trained on Kayapó minimal pairs.

Another challenge is the prevalence of complex onset clusters like /kra/, /mro/, and /ʔm/. For example, the word kra means “child”, but an ASR system unfamiliar with Kayapó might transcribe it as “krah” or split it into two syllables. Speechyou’s training data includes thousands of such clusters extracted from natural speech, enabling accurate segmentation.

Dialectal differences add another layer. The Gorotire dialect is the most widely used in literacy, but Xikrin speakers pronounce certain vowels more openly, and Kararaô speakers retain older consonant combinations. Speechyou allows users to select their dialect from a dropdown, which activates dialect-specific acoustic parameters.

Use cases: from oral history to subtitles

Kayapó communities increasingly use digital media to share their culture. A Kayapó radio program might be recorded and then transcribed for archival purposes. A documentary producer may need to subtitle an interview with a Kayapó elder for a global audience. Researchers studying the language need time-aligned transcriptions for phonetic analysis. With Speechyou, all these tasks become efficient: simply upload an audio or video file, select “Kayapó” as the language, and within minutes receive a timestamped transcript in the Latin-based orthography. The transcript can be exported as plain text, SRT, or VTT subtitles.

How Speechyou supports Kayapó

Our model was trained in collaboration with Kayapó language specialists, using ethical data collection protocols. The dataset covers the five main dialects and includes both read speech and spontaneous conversation. Accuracy tests show over 90% word error rate on clean recordings and acceptable performance on field recordings with moderate background noise. For users needing higher accuracy, the platform offers a simple correction interface where they can adjust misrecognised words.

Get started with Kayapó transcription today

Whether you are a linguist, an educator, or a Kayapó community member, Speechyou makes it easy to transcribe, subtitle, and preserve the Mebêngôkre language. No pricing per minute: transcription is unlimited in the Solo plan. Upload your Kayapó audio now and join the effort to keep this Amazonian language alive in the digital age.