AfroEdge

01 · The problem

Many African artists, particularly in Nigeria, lack access to professional recording studios. The goal: put a full music-creation workflow — AI generation through track editing — directly on their phones.

02 · What I built

The Flutter app end to end: a generation flow with custom Yoruba lyrics, genre and instrument pickers, and 30s–4min output control; an in-app studio editor with beat-level FFmpeg manipulation and WebView waveform components; plus integration with a FastAPI backend orchestrating on-demand RunPod GPUs for ACE-Step 2 inference. I also curated the Yoruba-language training data for fine-tuning.

03 · The impact

The POC demonstrated end-to-end generation — lyrics to playable audio — and secured client approval for full-scale product development.

What I did

• Built the Flutter app for AI music generation (30s–4min songs) directly from mobile devices.
• Developed an in-app studio editor with beat-level manipulation, sound swapping, and segment rearrangement using FFmpeg.
• Implemented the full generation workflow UI: Yoruba lyrics, multi-select genres and instruments, adjustable duration.
• Connected to a FastAPI backend orchestrating on-demand RunPod GPUs for ACE-Step 2 inference.
• Curated Yoruba-language training data for fine-tuning the model on African musical styles.
• Delivered a successful POC that secured client approval for full-scale development.