VOICE · MODEL: WHISPER-LARGE-V3-TURBO STATUS · OPERATIONAL
VOICE · MODEL: WHISPER-LARGE-V3-TURBO 2026-05-04

A Whisper that actually speaks Moroccan.

Upload Darija, French or Arabic audio. We transcribe with Whisper-large-v3, normalize Darija spelling variants, and give you clean, searchable text.

How it works
01

Capture or upload

Record directly in your browser, or upload an MP3, WAV, M4A, or WebM file (max 25MB).

02

Whisper transcription

We run it through Whisper-large-v3-turbo on Groq — typically 50× faster than real-time.

03

Darija normalization

Our post-processor maps spelling variants (wash/wesh/wèsh) to a canonical form. This is the moat.

04

Clean transcript

Searchable text with timestamps, language detection, and segment-level breakdown.

Try it now
audio.webm
Detected language: Duration:
Normalized (Darija → canonical)
Raw Whisper output
Built for

Call centers

Transcribe customer calls in Darija. Search by content. Tag sentiment. Compliance archive.

Voice-first apps

Let users speak instead of type. Critical for low-literacy users, field workers, accessibility.

Media & podcasts

Auto-transcribe Moroccan podcasts, YouTube shows, news interviews. Generate subtitles.

Government & NGO

Transcribe community consultations, parliamentary debates, citizen feedback.