AI Audio Restoration Prototype
Budget: $100 – $200 USD
I need an AI-savvy audio engineer to build a lightweight, working prototype that cleans and digitises old family recordings. For this first phase, the tool only has to handle cassettes, CDs and VHS transfers, yet it should also accept existing digital files (MP3, WAV, MP4) and even the occasional MiniDisc rip so we can test edge cases.
Core tasks
• Set up an end-to-end enhancement pipeline in Python using Whisper, ffmpeg or similar libraries.
• Apply noise reduction and basic speech isolation, then run speech-to-text to create synced transcripts.
• Output two artefacts per file:
– Enhanced audio (MP3 or WAV)
– Text transcript (JSON or SRT)
• Provide a simple command-line utility or bare-bones web page that lets me batch-upload sources and download results.
• Keep the code tidy and well commented so my in-house team can hook it into our cloud workflow later.
Success looks like pressing “go,” watching a small test set run end-to-end, and receiving cleaner audio plus readable transcripts without manual tweaking. If the prototype proves reliable, we'll expand its features and interface in a separate phase.
Core tasks
• Set up an end-to-end enhancement pipeline in Python using Whisper, ffmpeg or similar libraries.
• Apply noise reduction and basic speech isolation, then run speech-to-text to create synced transcripts.
• Output two artefacts per file:
– Enhanced audio (MP3 or WAV)
– Text transcript (JSON or SRT)
• Provide a simple command-line utility or bare-bones web page that lets me batch-upload sources and download results.
• Keep the code tidy and well commented so my in-house team can hook it into our cloud workflow later.
Success looks like pressing “go,” watching a small test set run end-to-end, and receiving cleaner audio plus readable transcripts without manual tweaking. If the prototype proves reliable, we'll expand its features and interface in a separate phase.