Secoya Speech to Text: Bringing an Amazonian Language into the Digital Age

The Secoya Language and Its Speakers

Secoya (also known as Sikoya, Paicoca, or Airo) is a Tucanoan language spoken by small communities along the Aguarico and Napo rivers in Ecuador and Peru. With fewer than 500 fluent speakers, Secoya is classified as endangered. Most speakers are bilingual in Spanish, and the language is primarily oral. Only a limited amount of written material exists, mostly produced by linguists and missionary organizations. The standard Latin-based orthography includes special characters such as ɨ, ñ, and diacritics for nasalization and tones.

Why Accurate Speech-to-Text for Secoya Matters

Preserving an endangered language requires capturing not just words, but the exact pronunciation, tone, and context. Traditional transcription is slow and expensive; one hour of audio can take a trained linguist many hours to transcribe manually. Speechyou's AI provides a scalable solution for transcribing Secoya audio into text with high fidelity. This enables:

• Language documentation – rapid conversion of oral histories, songs, and interviews into searchable digital archives.
• Educational materials – creation of bilingual subtitles for schools and community learning.
• Media accessibility – adding Secoya captions to videos for wider distribution.
• Research – phonetic and grammatical analysis of natural speech data.

Specific Transcription Challenges in Secoya

Tonal Contrasts

Secoya uses two levels of tone (high and low) that are lexically distinctive. For example, the word 'mɨa' with a high tone can mean 'speech', while with a low tone it might mean a different concept. Most commercial ASR systems are not built to handle tone languages, but Speechyou's model includes pitch tracking that differentiates tonal meanings. Users can also mark tone in the transcription output if their orthography requires it.

Nasal Vowels and Vowel Length

Secoya features phonemic nasal vowels (e.g., ɨ̃, ã) and contrastive vowel length. These are often missed by generic ASR, leading to erroneous transcriptions. Speechyou's acoustic model is sensitive to these cues because it was trained on a diverse set of languages including Amazonian ones. The system will reliably output the correct diacritics.

Data Scarcity

With only a few hundred speakers, Secoya is a true low-resource language. Most ASR engines simply ignore such languages. Speechyou takes a different approach: it uses a base multilingual model fine-tuned with even a small amount of target audio. Users can upload their own recordings to improve accuracy incrementally. This means the more people use it, the better it becomes.

Real-World Use Cases

• Oral history preservation: Elders narrate creation myths in Secoya. The audio is transcribed and subtitled in Secoya and Spanish, then uploaded to an online archive accessible to descendants.
• Bilingual education: Teachers use Speechyou to transcribe classroom dialogues. They export SRT files to add subtitles to video lessons, helping children learn both Secoya and Spanish literacy.
• Community radio: A Secoya radio station records interviews and news. Speechyou transcribes each broadcast, enabling the creation of a written newsletter for those who prefer reading.
• Legal and health interpreters: When Secoya speakers interact with government officials, the conversations are recorded and transcribed by Speechyou, ensuring accurate records.

How Speechyou Helps

Speechyou offers a simple web interface where users upload audio or video files (or record live). The AI processes the speech in real time or on upload, producing text that can be edited and exported in multiple formats. For Secoya, the key features are:

• Tone-aware transcription – the model identifies high and low tones and outputs them according to the user's preference.
• Subtitle generation – create SRT and VTT files directly from the transcribed text.
• Continuous learning – each transcription can be submitted as training data to improve future accuracy.
• 100+ languages in one account – no need to switch tools for multilingual projects that mix Secoya and Spanish.

By providing reliable, user-friendly speech-to-text for Secoya, Speechyou empowers speakers and linguists to document, teach, and share this unique Amazonian language. Whether you are a mother recording lullabies for her grandchildren or a researcher analyzing verb conjugations, Speechyou makes accurate transcription accessible to everyone.