[PhotoBooth] mode with image persistence along side audio file. Also, streaming capabilities, setup script, readme updates, and narrator prompt update

.env.example

@@ -0,0 +1,8 @@
+# Required variables:
+export OPENAI_API_KEY=
+export ELEVENLABS_API_KEY=
+export ELEVENLABS_VOICE_ID=
+
+# Optional variables:
+export ELEVENLABS_STREAMING=
+export PHOTOBOOTH_MODE=

.gitignore

@@ -2,4 +2,5 @@
 /venv
 /narration
 /frames/*
-!/frames/.gitkeep
+!/frames/.gitkeep
+.trunk

README.md

@@ -1,8 +1,9 @@
-# David Attenborough narrates your life. 
+# David Attenborough narrates your life
 
 https://twitter.com/charliebholtz/status/1724815159590293764
 
 ## Want to make your own AI app?
+
 Check out [Replicate](https://replicate.com). We make it easy to run machine learning models with an API.
 
 ## Setup
@@ -20,26 +21,57 @@ Then, install the dependencies:
 
 Make a [Replicate](https://replicate.com), [OpenAI](https://beta.openai.com/), and [ElevenLabs](https://elevenlabs.io) account and set your tokens:
 
-```
+```bash
 export OPENAI_API_KEY=<token>
 export ELEVENLABS_API_KEY=<eleven-token>
 ```
 
 Make a new voice in Eleven and get the voice id of that voice using their [get voices](https://elevenlabs.io/docs/api-reference/voices) API, or by clicking the flask icon next to the voice in the VoiceLab tab.
 
-```
+```bash
 export ELEVENLABS_VOICE_ID=<voice-id>
 ```
 
 ## Run it!
 
 In on terminal, run the webcam capture:
+
 ```bash
 python capture.py
 ```
+
 In another terminal, run the narrator:
 
 ```bash
 python narrator.py
 ```
 
+## Options
+
+### Setup
+
+#### Script
+
+Alternative to running the [Setup](#setup) commands above individually, one can use the `setup.sh` script to facilitate getting the two required shell envs ready to rock.
+
+_Note: will have to run `source source venv/bin/activate` afterwards to activate the virtual env._
+
+#### Dotenv
+
+One can set the environment variables via the `.env` file, which is read every time the process starts. It is recommended to copy the `.env.example` file and rename to `.env`.
+
+### Streaming
+
+If you would like the speech to start quicker via a streaming manner set the environment variable to enable. The concession is that the audio and corresponding image is not saved in the `/narration` directory.
+
+```bash
+export ELEVENLABS_STREAMING=true
+```
+
+### PhotoBooth
+
+The default behavior of this app will continually analyze images. If you would like to use in a mode more similar to a photo booth, set the environment variable. In this mode, the image will only be analyzed when the spacebar key is pressed.
+
+```bash
+export PHOTOBOOTH_MODE=true
+```

capture.py

@@ -1,8 +1,9 @@
-import cv2
+import os
 import time
-from PIL import Image
+
+import cv2
 import numpy as np
-import os
+from PIL import Image
 
 # Folder
 folder = "frames"
@@ -30,7 +31,7 @@
         # Resize the image
         max_size = 250
         ratio = max_size / max(pil_img.size)
-        new_size = tuple([int(x*ratio) for x in pil_img.size])
+        new_size = tuple([int(x * ratio) for x in pil_img.size])
         resized_img = pil_img.resize(new_size, Image.LANCZOS)
 
         # Convert the PIL image back to an OpenCV image

narrator.py

@@ -1,16 +1,31 @@
-import os
-from openai import OpenAI
 import base64
-import json
-import time
-import simpleaudio as sa
 import errno
-from elevenlabs import generate, play, set_api_key, voices
+import os
+import shutil
+import time
+
+from dotenv import load_dotenv
+from elevenlabs import generate, play, set_api_key, stream
+from openai import OpenAI
+from pynput import (  # Using pynput to listen for a keypress instead of native keyboard module which was requiring admin privileges
+    keyboard,
+)
+
+# import environment variables from .env file
+load_dotenv()
 
 client = OpenAI()
 
 set_api_key(os.environ.get("ELEVENLABS_API_KEY"))
 
+# Initializes the variables based their respective environment variable values, defaulting to false
+isStreaming = os.environ.get("ELEVENLABS_STREAMING", "false") == "true"
+isPhotoBooth = os.environ.get("PHOTOBOOTH_MODE", "false") == "true"
+
+script = []
+narrator = "Sir David Attenborough"
+
+
 def encode_image(image_path):
     while True:
         try:
@@ -24,17 +39,30 @@ def encode_image(image_path):
             time.sleep(0.1)
 
 
-def play_audio(text):
-    audio = generate(text, voice=os.environ.get("ELEVENLABS_VOICE_ID"))
+def play_audio(text, dir_path=None):
+    audio = generate(
+        text,
+        voice=os.environ.get("ELEVENLABS_VOICE_ID"),
+        model="eleven_turbo_v2",
+        stream=isStreaming,
+    )
 
-    unique_id = base64.urlsafe_b64encode(os.urandom(30)).decode("utf-8").rstrip("=")
-    dir_path = os.path.join("narration", unique_id)
-    os.makedirs(dir_path, exist_ok=True)
+    if isStreaming:
+        # Stream the audio for more real-time responsiveness
+        stream(audio)
+        return
+
+    # Save the audio file to the directory
     file_path = os.path.join(dir_path, "audio.wav")
 
     with open(file_path, "wb") as f:
         f.write(audio)
 
+    # Copy the image analyzed to the same directory as the audio file
+    image_path = os.path.join(os.getcwd(), "./frames/frame.jpg")
+    new_image_path = os.path.join(dir_path, "image.jpg")
+    shutil.copy(image_path, new_image_path)
+
     play(audio)
 
 
@@ -43,7 +71,10 @@ def generate_new_line(base64_image):
         {
             "role": "user",
             "content": [
-                {"type": "text", "text": "Describe this image"},
+                {
+                    "type": "text",
+                    "text": f"Describe this image as if you are {narrator}",
+                },
                 {
                     "type": "image_url",
                     "image_url": f"data:image/jpeg;base64,{base64_image}",
@@ -59,8 +90,8 @@ def analyze_image(base64_image, script):
         messages=[
             {
                 "role": "system",
-                "content": """
-                You are Sir David Attenborough. Narrate the picture of the human as if it is a nature documentary.
+                "content": f"""
+                You are {narrator}. Narrate the picture of the human as if it is a nature documentary.
                 Make it snarky and funny. Don't repeat yourself. Make it short. If I do anything remotely interesting, make a big deal about it!
                 """,
             },
@@ -73,30 +104,71 @@ def analyze_image(base64_image, script):
     return response_text
 
 
-def main():
-    script = []
+def _main():
+    global script
+
+    # path to your image
+    image_path = os.path.join(os.getcwd(), "./frames/frame.jpg")
+
+    dir_path = None
+    if not isStreaming:
+        # create a unique directory to store the audio and image
+        unique_id = base64.urlsafe_b64encode(os.urandom(30)).decode("utf-8").rstrip("=")
+        dir_path = os.path.join("narration", unique_id)
+        os.makedirs(dir_path, exist_ok=True)
+
+        # copy the image to the directory
+        new_image_path = os.path.join(dir_path, "image.jpg")
+        shutil.copy(image_path, new_image_path)
+        image_path = new_image_path
+
+    # getting the base64 encoding
+    base64_image = encode_image(image_path)
+
+    # analyze the image
+    print(f"👀 {narrator} is watching...")
+    analysis = analyze_image(base64_image, script=script)
 
+    print(f"🎙️ {narrator} says:")
+    print(analysis)
+
+    # generate and play audio
+    play_audio(analysis, dir_path)
+
+    script = script + [{"role": "assistant", "content": analysis}]
+
+
+def main():
     while True:
-        # path to your image
-        image_path = os.path.join(os.getcwd(), "./frames/frame.jpg")
+        if isPhotoBooth:
+            pass
+        else:
+            _main()
+
+            # wait for 5 seconds
+            time.sleep(5)
+
 
-        # getting the base64 encoding
-        base64_image = encode_image(image_path)
+def on_press(key):
+    if key == keyboard.Key.space:
+        # When space bar is pressed, run the main function which analyzes the image and generates the audio
+        _main()
 
-        # analyze posture
-        print("👀 David is watching...")
-        analysis = analyze_image(base64_image, script=script)
 
-        print("🎙️ David says:")
-        print(analysis)
+def on_release(key):
+    if key == keyboard.Key.esc:
+        # Stop listener
+        return False
 
-        play_audio(analysis)
 
-        script = script + [{"role": "assistant", "content": analysis}]
+# Create a listener
+listener = keyboard.Listener(on_press=on_press, on_release=on_release)
 
-        # wait for 5 seconds
-        time.sleep(5)
+# Start the listener
+listener.start()
 
+if isPhotoBooth:
+    print(f"Press the spacebar to trigger {narrator}")
 
 if __name__ == "__main__":
     main()

requirements.txt

@@ -28,6 +28,8 @@ pure-eval==0.2.2
 pydantic==2.4.2
 pydantic_core==2.10.1
 Pygments==2.16.1
+pynput==1.7.6
+python-dotenv==1.0.0
 requests==2.31.0
 simpleaudio==1.0.4
 six==1.16.0
@@ -38,4 +40,4 @@ traitlets==5.13.0
 typing_extensions==4.8.0
 urllib3==2.0.7
 wcwidth==0.2.10
-websockets==12.0
+websockets==12.0

setup.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+# create a virtual environment
+python3 -m pip install virtualenv
+python3 -m virtualenv venv
+
+# source the virtual environment to install dependencies
+source venv/bin/activate
+
+# install the dependencies
+pip install -r requirements.txt
+
+echo -e "\n\n\nSetup complete. Run $(source venv/bin/activate) to activate the virtual environment.\n\nAlso, please ensure your environment variables are set correctly in the .env file."