Error loading ffprobe when deploying onto streamlit cloud

Hi All,

New to using streamlit, and really enjoying how easy it is to get projects running.

I have a project I am building out to create videos from text using several of the OpenAI API’s (DALLE, TTS, 3.5T) and have made it work locally with no problems. However, when deploying to streamlit cloud, I receive the error “No such file or directory: ‘ffprobe’” although I have ffmpeg in my packages.txt and ffmpeg-python==0.2.0 ffprobe==0.5 in my requirements.txt. Anyone experience a problem like this and know a solution?

Here is the repo: Text-to-Video/videoCreation at main · detroitnatif/Text-to-Video · GitHub , and I call the ffmpeg subprocess in video.py.

Thanks!

Hi @detroitnatif

Have you tried rebooting the app and see if that may help trigger the installation of the packages.

Yes. I actually see it being installed in the logs here:
Stored in directory: /tmp/pip-ephem-wheel-cache-mg566x2y/wheels/ba/39/54/c8f7ff9a88a644d3c58b4dec802d90b79a2e0fb2a6b884bf82

Building wheel for ffprobe (setup.py): started

Building wheel for ffprobe (setup.py): finished with status ‘done’
Stored in directory: /tmp/pip-ephem-wheel-cache-mg566x2y/wheels/77/95/42/ec9cdac2db6ff5c7fafb2f430715681009ddd232c7704f517f

Successfully built playsound ffprobe

Installing collected packages: wcwidth, pytz, pydub, pure-eval, ptyprocess, playsound, ffprobe

even with this installed, I get the original error.

I solved this problem by using ffmpeg python wrapper rather than calling the binary. I made this helper function and called it in place of subprocess.run(ffmpeg).

def merge_audio_video(avi_video_path, full_narration_path, output_file_path):
“”"
Merge AVI video and MP3 narration into a final MP4 video using ffmpeg-python.

Parameters:
- avi_video_path: Path to the AVI video file.
- full_narration_path: Path to the MP3 narration audio file.
- output_file_path: Path for the output MP4 video file.
"""
try:
    # Input streams
    video_input = ffmpeg.input(avi_video_path)
    audio_input = ffmpeg.input(full_narration_path)
    
    # Merge video and audio with the specified output format and codec options
    (
        ffmpeg
        .output(video_input['v'], audio_input['a'], output_file_path, vcodec='copy', acodec='aac', strict='experimental', shortest=None)
        .run(overwrite_output=True)
    )
    print("Video processing completed successfully.")
except ffmpeg.Error as e:
    print(f"ffmpeg command failed: {e.stderr.decode()}")