Build products that understand Ethiopian speech.
Transcribe Amharic, Afaan Oromo, Tigrinya, and Ethiopian English with structured speaker diarization, sub-second timestamps, and verifiable evidence you can trace directly to the recording.
"ሰላም ጤና ይስጥልኝ ወ/ሮ ትዕግስት፣ በሱቅዎ ውስጥ ዲጂታል ክፍያዎችን መጠቀም ከጀመሩ ምን ያህል ጊዜ ሆነዎት?"
"Greetings Ms. Tigist, how long has it been since you began accepting digital payments in your shop?"
"ሰላም፣ ቴሌብር እና ሲቢኢ ብር መጠቀም ከጀመርን አሁን ወደ አንድ አመት ከስድስት ወር ሆኖናል።"
"Hello, it has been about one year and six months since we began using Telebirr and CBE Birr."
"ደንበኞች በቴሌብር ለመክፈል ፍላጎት አላቸው፣ ነገር ግን አንዳንዴ ኔትወርክ ሲቋረጥ ክፍያው ዘግይቶ መድረሱ ያሳስበናል።"
"Customers are keen to pay via Telebirr, but occasionally network dropouts delay the confirmation message, which worries us."
"እሺ፣ ይህ ደንበኞችን ወደ ጥሬ ገንዘብ እንዲመለሱ ያደርጋቸዋል?"
"I see, does this make customers revert to physical cash?"
"አዎ፣ አብዛኛው ሰው በጊዜ ማረጋገጫ ካላገኘ ጥሬ ገንዘብ ያወጣል። ይሁን እንጂ ቅዳሜና እሁድ የክፍያው ፍሰት በጣም ፈጣን ነው።"
"Yes, most people pull out cash if they do not receive an instant notification. However, on weekends transaction flow is usually quite smooth."
"ደንበኞች በቴሌብር ለመክፈል ፍላጎት አላቸው፣ ነገር ግን አንዳንዴ ኔትወርክ ሲቋረጥ ክፍያው ዘግይቶ መድረሱ ያሳስበናል።"
Your first API request
Copy the request for your preferred environment or test the live response contract right in the browser.
Synchronous transcription for recordings under 120 seconds. Returns complete text, speakers, and evidence immediately.
Developer Sandbox Allowance
Every account starts with complete access to speech intelligence features.
Built for real integrations
A production audio engine built for engineering teams, with concrete guarantees and unambiguous contracts.
Ethiopian Language Specialization
Tuned on thousands of hours of spoken Amharic (አማርኛ), Afaan Oromoo, Tigrinya (ትግርኛ), and Ethiopian English. Native Ge'ez script encoding and resilience to heavy code-switching in retail and technical conversations.
Speaker Diarization & Evidence Grounding
Separates overlapping speakers and assigns consistent identifiers (spk_01, spk_02). Every quote and extracted entity anchors back to millisecond audio offsets for auditable human verification.
Dual Execution Models
Files under 120s return an immediate 200 OK response. Long files (up to 4 hours) queue with 202 Accepted, supported by status polling and HMAC-signed webhook delivery with replay attack protection.
Idempotency & Predictable Economics
Support for Idempotency-Key headers prevents duplicate charges during network retries. Zero audio retention mode available. Audio billed in 1-second increments with no charge for failed calls.
Operational Specifications
Request Contract Inspector
Inspect authentic request fixtures, response envelopes, and error signatures.