sirdavid

Have you ever wanted David Attenborough's voice to describe the contents of a webcam photo in pseudo-real time over a secure websocket? Why not indeed!

Live demo up at https://sirdavid.rickt.dev

index.html:

• takes a snapshot of the browser's webcam
• uploads it over secure websocket to a mini backend

main.py:

• sets up a secure websocket listener
• looks for a .png sent over the websocket
• if an image arrives, saves it locally and in a GCP bucket
• decodes the image, has it described by OpenAI's gpt-4o model in a snarky David Attenborough manner
• generates audio file using a custom ElevenLabs David Attenborough voice i created
• sends the URL of the audio file to the browser
• plays the audio in the browser

Environment Variables:

You'll need to set:

• ELEVENLABS_API_KEY api key for ElevenLabs
• OPENAI_API_KEY api key for OpenAI
• SIRDAVID_APIGW URI pointing to (in this case) my Cloudflare AI Gateway
• SIRDAVID_BUCKET name of GCP bucket to store images, text analyses & audio files
• SIRDAVID_PORT port for the websocket listener
• SIRDAVID_SERVICEACCOUNT_JSON path to service account JSON file for auth to GCP bucket
• SIRDAVID_SSL_CERT path to the SSL certificate for the secure websocket
• SIRDAVID_SSL_KEY path to the SSL key for the secure websocket certificate
• SIRDAVID_VOICE string containing the voice ID of the ElevenLabs voice

Notes

• You will have to make your own ElevenLabs David Attenborough voice, as I can't share mine