Contact person: Business Client

Phone:Show

Email:Show

Location: Port Harcourt, Nigeria

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I’m looking for a small utility—CLI or lightweight desktop/web app—that lets me point to an MP3 stored locally, hits Whisper to create the transcript, then feeds both the audio and transcript straight into Montreal Forced Aligner. The result I need is a JSON file containing word- and phoneme-level timing that is immediately offered as a direct download once processing finishes.

Workflow I expect
1. Select local MP3 (no other formats required).
2. Whisper runs automatically, producing a transcript.
3. Transcript and original MP3 are passed to MFA for alignment.
4. MFA output is converted into a clean JSON structure (one entry per phoneme with start/end offsets).
5. Browser (or CLI) prompts an instant download of that JSON.

I’m fine with Python—PyTorch Whisper bindings and MFA’s command-line tools make sense—but use whatever stack gets the job done reliably on Windows and macOS. Keep setup friction low: a single [login to view URL] or installer script is ideal.

Deliverables
• Source code with clear setup/usage instructions
• Sample JSON generated from a test MP3 so I can confirm format
• Short read-me explaining how to swap models or tweak Whisper/MFA parameters

Acceptance
If I can process an MP3 from my laptop and receive a JSON download matching MFA’s timing within ±20 ms, the project is complete." (client-provided description)