korymath/justforlaughs

Just For Laughs

Setup

# Get the code
git clone https://github.com/korymath/justforlaughs;
cd justforlaughs/;
python3 -m venv venv;
source venv/bin/activate;

# Install requirements
pip install --upgrade pip;
pip install -r requirements.txt;

Run

# Run the web app
python app.py;

# Run the standalone laughter detector
python segment_laughter.py --input_audio_file example_audio.wav;

# It should output that it found 1 laugh in the example, save just the laugh cropped from the input, and the time window when laugh happened.

# Example:
# found 1 laughs.
# [{'filename': 'output/laugh_0.wav', 'start': 2.6453333333333333, 'end': 5.261913043478261}]

Fixes

For running on mac, need to ensure that the libsndfile is correclty pathed, ref: bastibe/python-soundfile#310

env DYLD_LIBRARY_PATH="/opt/homebrew/lib:$DYLD_LIBRARY_PATH" python app.py

Credits

Client-sider in-browser detection from:

• https://github.com/tensorflow/tfjs-models/tree/master/speech-commands - LICENSE

Laughter detection model from:

• https://github.com/jrgillick/laughter-detection - LICENSE

Audio interface and recording adapted from:

• https://github.com/mattdiamond/Recorderjs - LICENSE
• https://github.com/addpipe/simple-recorderjs-demo - No LICENSE, Blog post

justforlaughs's Issues

Add UI elements to plot the soundwave and laughter segments

It would be great and fairly easy to plot the soundwave (as an image, using the recording that you process)

and highlight in red the laughter bit.

Ensure requests are well handled when the server is not available

I tried to run it and got it to record but it did not detect anything at first.
Then I saved the file and clicked again and it ran the detection.

 POST https://justforlaughs-yww6svdp4q-nn.a.run.app/ 503
(anonymous) @ app.js:165
VM27:1 Uncaught SyntaxError: Unexpected token 'S', "Service Unavailable" is not valid JSON
    at JSON.parse (<anonymous>)
    at xhr.onload (app.js:154:35)
xhr.onload @ app.js:154
load (async)
(anonymous) @ app.js:152
app.js:165          POST https://justforlaughs-yww6svdp4q-nn.a.run.app/ 503
(anonymous) @ app.js:165
VM28:1 Uncaught SyntaxError: Unexpected token 'S', "Service Unavailable" is not valid JSON
    at JSON.parse (<anonymous>)
    at xhr.onload (app.js:154:35)
xhr.onload @ app.js:154
load (async)
(anonymous) @ app.js:152
app.js:165          POST https://justforlaughs-yww6svdp4q-nn.a.run.app/ 503
(anonymous) @ app.js:165
VM29:1 Uncaught SyntaxError: Unexpected token 'S', "Service Unavailable" is not valid JSON
    at JSON.parse (<anonymous>)
    at xhr.onload (app.js:154:35)
xhr.onload @ app.js:154
load (async)
(anonymous) @ app.js:152
app.js:165          POST https://justforlaughs-yww6svdp4q-nn.a.run.app/ 503
(anonymous) @ app.js:165
VM30:1 Uncaught SyntaxError: Unexpected token 'S', "Service Unavailable" is not valid JSON
    at JSON.parse (<anonymous>)
    at xhr.onload (app.js:154:35)

Add UI instructions for how long things are going to take.

Remove the filename and `Save to disk`

These are not used and only complexify things.

Add visualization so that the user knows audio is coming in

Example audio canvas can be seen here:

• https://github.com/mdn/web-dictaphone
• https://musiclab.chromeexperiments.com/spectrogram/

Add UI instructions for how long things are going to take.

Update the UI to include details on what to do.

This is just a laughter detector.

Instructions:

1. click record (you might need to allow access to your mic), and record a short sample (2 seconds) of you laughing / talking / etc
2. click stop, listen to your recording, if you are happy with it, click Detect Laughter
3. After about 10 seconds or so, the laughter detector will respond with a number of laugh instances, and when in the sample they were.