Recherche avancée

Sur d’autres sites (3500)

• Is -vfilter not available on Debian 5 ?

  3 octobre 2011, par Kor

  I'm getting problems trying to make my website (done in PHP) work online.
  At a specific time, I need to upload a video and convert it, also scaling its dimensions. I use a call like this one :

  ffmpeg -y -i '/path/to/video.mp4' -f flv -ab 56 -ar 22050 -r 20 -vfilter "scale=704:-1" '/path/to/new/video.flv'

  It works offline on my computer (Ubuntu 10.10), but it doesn't do the same online (Debian Lenny 5.0). The response I get from the server is :

  [Mon Oct 03 15:48:56 2011] [error] [client 81.184.6.124] /servers/path/to/ffmpeg: unrecognized option '-vfilter'

  I have also tried with '-vf', but it just doesn't work at all. So my question is, am I doing it wrong, or is it that Debian unables video filters for some reason ?

  I give you some info about this server :

  Debian 5.0
  
  PHP: 5.2.6
  
  FFmpeg r11872+debian_0.svn20080206-18+lenny1
  
  libavutil   3212800
  
  libavcodec  3355136
  
  libavformat 3409664
  
  libavdevice 3407872

  EDIT : Oh my, I just saw it. No libavfilter installed. I'll tell you what their support tells me in a couple hours.

• Why does OpenAI Whisper return "None" ?

  29 septembre 2024, par SHDDSFFDSDSAF

  I'm trying to transcribe an MP3 file using OpenAI’s Whisper model, but the transcriptions.create() method consistently returns None. I’ve tried different MP3 files, converted them to WAV, updated the OpenAI library, and added error handling, but I still can’t figure out the issue.

  


  Here’s my code :

  


  from openai import OpenAI

client = OpenAI(
    api_key="MYAPIKEY"
)

audio_file = open("speech.mp3", "rb")
transcript = client.audio.transcriptions.create(
    file=audio_file,
    model="whisper-1",
    response_format="verbose_json",
    timestamp_granularities=["segment"]
)

print(transcript)


  


  I’ve confirmed that :

  


  

  • The API key is valid.

  


  • I’m using OpenAI Python library version X.X.X.

  


  • I’ve tested both MP3 and WAV formats.

  


  • The MP3 file is valid (checked using ffmpeg).

  


  


  The response always returns None without any exceptions. Any ideas on what could be going wrong ?

  


• How to encode three streams in live from a CloudStorage ?

  18 mars 2021, par Damien Bertholet

  I search a lot and doesn't find anyone posted that question. If I'm wrong, I'm sorry and let me know the link please.

  


  Here is what I would like to do. I have an application that can record his screen as well as his computer's microphone and system sound. I save the three pieces in separate files which I then send to my server.

  


  Then I write these files directly in Azure strorage, live. Once the recording is finished, it may not be fully uploaded yet, I would like that we can already listen to what has been uploaded.

  


  So this I would like my server to start three streams to Azure Storage (it's an easy part), put those three streams in ffmpeg and output me an output that I can stream in HTTP response.

  


  The processing will only merge the two audio tracks and re-encode the audio in aac (it may be in opus) but I would not touch the video (which is in h264).

  


  Also, and it seems this is a very important point, the video is in mp4 format.

  


  Does this sound theoretically feasible ? If so, any hint to helping me ?

  


  Thereafter, I will have a process that will transform the three files into a ready-to-use mp4 file so that I don't have to encode every time the file is requested. But I see no stakes for this part there.