Logo Lanfrica

wouapit999/benin-translator-mvp

Domain:

natural language processing

Record type:

software
Creator:
wou
Host:
Dentons KMN Translation Benin file # Benin Languages → English Translator (MVP) **Confirmed deployment choices:** - **Azure region:** West Europe - **Max video duration:** 12 minutes - **Outputs:** Dual subtitles (original language **and** English). Also individual tracks. Cloud-first, public API for **text + speech** translation targeting: **Fon, Yoruba, Bariba (Baatɔnum), Dendi, Goun, Adja**. It processes **TikTok videos** (or uploads), extracts audio, performs **ASR**, translates to English (direct or via **French pivot**), and outputs **SRT/VTT subtitles** (dual + individual tracks) and optional **English audio**. --- ## Dual Subtitles Modes - **Individual tracks**: `video.orig.srt` / `video.en.srt` (and `.vtt` variants) - **Bilingual track**: one SRT where **line 1 = original**, **line 2 = English** (useful for players that only accept one subtitle file) You can choose which outputs to produce via the `outputs` array in the API request: e.g., `["srt", "vtt", "json", "audio_en", "dual"]`. --- # Benin Languages → English Translator (MVP) Cloud-first, public API for **text + speech** translation targeting: **Fon, Yoruba, Bariba (Baatɔnum), Dendi, Goun, Adja**. It processes **TikTok videos** (or uploads), extracts audio, performs **ASR**, translates to English (direct or via **French pivot**), and outputs **SRT/VTT subtitles** and optional **English audio**. --- ## ✨ Features (MVP) - Submit TikTok URL or upload MP4/MOV/WEBM. - Audio extraction & chunking (FFmpeg). - ASR: Azure Speech (where supported) or Whisper (fallback). - Translation orchestrator: - **Yoruba**: direct `yo → en`. - **Fon, Goun, Adja, Bariba, Dendi**: **pivot `X → fr → en`** initially. - Subtitle generation: **SRT/VTT**. - Optional **English TTS** and audio muxing. - Public URLs via Azure Blob Storage + CDN. > ⚠️ Legal/Policy: Process only user-provided content with consent. Provide takedown contact. Translations may contain errors—verify critical content. --- ## 🧱 Architecture (High-Level) Client → API (FastAPI on …