YOLOv8 Image Caption Voice Assistant

I am building an accessibility-focused web page that can detect everyday objects in a user-supplied image, generate a short caption, and immediately read that caption aloud through a Text-to-Speech engine. The flow is simple: upload → YOLOv8 object detection → BLIP caption generation (Hugging Face) → TTS playback.
Users will operate the site with a standard keyboard and mouse, so all actions should be triggerable by clear buttons and keyboard shortcuts.

Core requirements
• Integrate YOLOv8 for real-time detection of everyday household items, furniture, food, etc.
• Feed the detected regions into the BLIP model to create a concise, grammatically correct caption.
• Convert that caption into spoken output using any reliable TTS library; natural pronunciation and minimal latency are more important than a “human-like” tone.
• Host everything inside a lightweight, responsive front-end (plain HTML/JS or a framework you prefer) with one upload area and a play / replay button.
• Ensure the entire pipeline runs automatically after the image is chosen, with visual progress indicators and error handling.

Deliverables
1. Clean, documented source code (backend + front-end).
2. Instructions to deploy locally and on a standard VPS.
3. A short screen-capture demo proving the end-to-end workflow.

I value clear comments, concise dependencies, and accessibility-friendly design decisions (e.g., ARIA labels, large buttons). If you have prior work combining computer vision, Hugging Face models, and TTS, I’d love to see a sample.