Integrate production speech-to-text, speaker diarization, and semantic audio search into your applications. Sub-second latency, verified webhooks, and official Python & TypeScript SDKs.
Instant API key generation · 45 free minutes · SDKs for Python & TypeScript
import { Andebet } from "@andebet/sdk";
const client = new Andebet({ apiKey: process.env.ANDEBET_API_KEY });
const result = await client.transcriptions.create({
audioUrl: "https://storage.andebet.ai/samples/addis_speech.wav",
language: "am", // Auto-detect or 'am' | 'om' | 'ti' | 'en'
diarize: true,
punctuate: true,
});
console.log(result.segments);
// Output: [{ speaker: "spk_1", start: 0.0, end: 4.2, text: "ሰላም ጤና ይስጥልኝ" }]Trying to spin up your own open-source Whisper or Conformer models for Amharic requires managing expensive GPU clusters, configuring custom Ge'ez tokenizers, writing custom audio chunking logic to handle files over 30 seconds, and tuning VAD (Voice Activity Detection) models for noisy African phone audio.
Worse, standard models lack speaker diarization and code-switching capabilities, leaving engineering teams to build complex post-processing microservices that take months to stabilize.
With the Andebet API, you replace that entire pipeline with a single REST call. Pass an audio URL, receive timestamped and diarized speech segments in sub-second response times, and scale effortlessly from 10 to 100,000 hours per month.
Intuitive REST semantics, structured JSON outputs, and comprehensive error schemas.
Submit audio files via URL or binary multipart payload. Configure language auto-detection, speaker separation, and Ge'ez punctuation normalization.
Poll job progress for batch recordings. Retrieves full word-level timestamps, speaker turn intervals, and language transition markers.
Query an entire collection of audio recordings in plain language. Returns grounded text passages with exact audio start and end timestamps.
Programmatically query consumed transcription hours, active rate limit headroom, and webhook delivery success rates.
Built for production systems running thousands of concurrent audio streams.
Secure asynchronous notifications sent to your backend upon job completion with signature verification headers.
Process multi-hour broadcasts and continuous audio streams effortlessly with automatic parallel chunking.
Fully typed libraries with autocomplete, built-in retry mechanisms, exponential backoff, and error handling.
Flexible payment options. Pay in USD via international credit cards or in Ethiopian Birr via local payment gateways.
Test endpoints, inspect raw JSON responses, and copy ready-to-run curl snippets before writing any code.
Enterprise dedicated instances deployed in your AWS, GCP, or Azure tenant for regulated data compliance.
“Integrating Andebet's Python SDK took us less than an hour. We now process thousands of customer service voice calls daily, extracting automated compliance metrics with 98% accuracy in Amharic.”
Technical answers regarding our speech API, SDKs, and infrastructure.
The API natively supports Amharic (including Fidel orthographic normalization), Afaan Oromo (Qubee Latin script), Tigrinya, and English. Our acoustic models automatically detect the spoken language and gracefully handle code-switching within the same recording.
Short audio clips (under 30 seconds) process synchronously in under 800ms. Long audio files (up to several hours) process asynchronously at approximately 15x real-time speed (e.g. an hour-long recording is transcribed, diarized, and punctuated in approximately 4 minutes).
When creating a transcription job via POST /v1/transcriptions, you can pass a webhook_url parameter. Once processing completes, Andebet sends an HTTP POST request containing the full transcript, speaker diarization segments, and confidence scores. Every webhook payload is signed with an HMAC SHA-256 signature in the X-Andebet-Signature header.
Yes. We maintain official client libraries for Python (pip install andebet) and TypeScript/Node.js (npm install andebet). Both libraries feature full TypeScript typings, automatic retries with exponential backoff, and helper methods for file uploads and webhook verification.
Audio is billed per second of processed audio, with no minimum file length penalties. Free accounts include 50 hours of complimentary processing. Paid tiers offer predictable volume-discounted rates starting at $0.006 per minute, with custom enterprise plans available for dedicated infrastructure.
Yes. For government entities, financial institutions, and enterprise clients with strict data sovereignty mandates, we provide dedicated containerized deployments and air-gapped VPC appliances that run within your AWS, Azure, or private cloud boundary.
Create your developer account, generate an API key, and make your first transcription call in under sixty seconds.