Pashayi is a Dardic language spoken by an estimated 400,000 people across eastern Afghanistan, primarily in the provinces of Laghman, Kunar, Nangarhar, and Kapisa, with smaller communities in Pakistan. It belongs to the Indo-Aryan branch of the Indo-European family, sharing features with other Dardic languages like Kashmiri and Shina. The language is notable for its rich system of retroflex consonants, a complex vowel inventory with length contrasts, and a stress accent that can differentiate words. Despite its relatively small speaker population, Pashayi is highly diverse, with four major dialect groups: Southeastern, Northeastern, Northwestern, and Southwestern. These dialects differ in phonology, lexicon, and even basic grammatical structures. The Latin script romanization used in Speechyou's model follows a consistent orthography that can represent all dialectal sounds, making it a practical tool for transcription and subtitle generation.

Accurate Pashayi speech-to-text is essential for several reasons. First, the language is under-documented, and automated transcription can accelerate linguistic research and language preservation. Second, media content in Pashayi — from local radio broadcasts to YouTube videos — lacks subtitles, limiting accessibility for Pashayi speakers who are hard-of-hearing or have low literacy. Third, NGOs and health organizations need to communicate health messages in Pashayi, and a text version ensures the information can be translated into other languages. Finally, educators and community leaders want to produce educational materials in Pashayi to foster mother-tongue literacy. Speechyou's AI fills this gap by providing a reliable, low-cost solution that works with minimal training data, thanks to its transfer learning from related languages.

The transcription challenges for Pashayi are significant. Its retroflex consonant series (e.g., /ʈ/, /ɖ/, /ɳ/) is acoustically similar to dental stops in many languages, requiring precise acoustic modeling. Vowel length is phonemic: /a/ vs /aː/ can change the meaning of a word. Additionally, the stress accent is not predictable and can be contrastive. Dialectal variation adds another layer of complexity; for example, the Southeastern dialect retains a pitch accent, while the Southwestern dialect has reduced vowels. Speechyou's model is trained on a carefully curated corpus of Pashayi speech from multiple dialects, with data augmentation to handle background noise, and uses a language model that incorporates dialect-specific n-grams. The result is a word accuracy exceeding 95% on clean audio and robust performance even in challenging acoustic environments.

Speechyou makes Pashayi transcription practical for everyday use. Users can upload audio or video files, and the system returns a time-stamped transcript in SRT or VTT subtitle format. The web interface allows editing and exporting. For native speakers, the transcription is in Latin script, which is widely used in digital communication by Pashayi speakers. The model also supports punctuation and capitalization, making the output ready for publication. Whether you are a researcher analyzing oral narratives, a content creator adding subtitles, or a community worker documenting health advice, Speechyou streamlines the process. And because the AI is cloud-based, you can access it from any device without specialized hardware. Pashayi is just one of the 100+ languages you can transcribe, but our dedicated model ensures that minority languages are not left behind in the AI revolution.