Recherche avancée

Sur d’autres sites (14181)

• We're Creating a New AI tool and Got Stuck on This 1 Step [closed]

  22 février 2024, par Daniel

  Thanks for helping with this question,

  


  Anybody who knows PHP, React js, FFmpeg, or Open AI API can help here :

  


  We are currently creating a tool with PHP laravel and React JS. It lets normal creators input a simple text prompt and they'll get a fully original short (30 video fast-paced video) for their channel. But we have an output problem :

  


  

  • On the output of the videos, there's supposed to be AI generated images that have cool animations in order to make everything more fast paced. But our developers are making animations that are super slow and not dynamic or interesting enough. What open source library/tech could we use to create stunning animations and more action ?! Any suggestions would help here,

  


  


  ---

  


  We are currently using an FFmpeg to create a video automatically for our users. The videos consists of [images, captions, voiceover, background music, and animations) all generated by our AI tool and put together.

  


  Here's what we want as a video (dynamic, fast paced, and many animations) : https://drive.google.com/file/d/1ONCczJh_Uk9m6oRDPGFQYjJPWUhy8hdV/view?usp=sharing

  


  Here's what we're getting : https://drive.google.com/file/d/1sMKGB88ouTHsEdahc0Zyt8rKEvZWhTco/view?usp=sharing

  


  Thanks for any help here.

  


• ffmpeg poor YUV to RGB in sws_scale() (was : ffmpeg blocky chroma decoding)

  12 septembre 2017, par noiseshaper

  I’m using ffmpeg-3.2.4-win32, static and dev versions from zeranoe page, to decode some h264 mp4 videos I have created with x264 encoder. The videos have lots of static zones, 2d videogame type.

  When I decode their frames using either ffmpeg.exe or avcodec_decode_video2() API function, I get poor chroma resolution, blocky and quite visible in the static zones of the video (edit : avcodec_decode_video2() returns correct YUV, the problem is the YUV->RGB conversion, see my own answer below).

  With ffmpeg.exe I get poor chroma only if I output the frames to png, but if I output them to jpg, chroma seems much better (edit : this is due to poor YUV to RGB conversion of ffmpeg’s sws_scale()).

  Here is an example of the same frame obtained from a mp4 video created with a static image, the output frame is zoomed in to clearly see the effect.

  https://drive.google.com/file/d/0B5KI1D-N1kHpV3lGaERJRjNkcms/view

  1 : Original frame before compressing.

  2 : YUV 4:2:0 conversion before compressing

  3 : Video output from MPC HC + LAV decoder. This is YUV decoded output.

  4 : Video frames decoded with ffmpeg.exe to jpeg, quality is mostly ok, just some jpeg macroblocking is visible.

  5 : Video frames decoded with ffmpeg.exe to png, very blocky chroma. This is RGB output, using avcodec_decode_video2() and sws_scale() conversion to RGB produces the same blocky chroma.

  ---

  The command lines for decoding the frames to jpg and png are these :

  ffmpeg -i testcase.mp4 -vframes 5 -qscale:v 2 output%03d.jpg

  ffmpeg -i testcase.mp4 -vframes 5 output%03d.png

  You can download mp4 and bat files that showcase the effect here :

  https://drive.google.com/drive/folders/0B5KI1D-N1kHpcUdGd2IyM2pXMTg

  ---

  I have tried with ffmpeg 3.2, but same thing happens.

• ffmpeg : avfilter's anull says "Rematrix is needed between stereo and 0 channels but there is not enough information to do it"

  12 juillet 2017, par kuanyui

  I’m trying to write a transcoder according to FFMpeg’s official example with ffmpeg 3.2.4 (official prebuild Win32), and try to transcode a video with stereo audio stream source (from avformat’s dshow).

  In the example code, which passes anull into avfilter_graph_parse_ptr() for audio stream, and "time_base=1/44100:sample_rate=44100:sample_fmt=s16:channels=2:channel_layout=0x3" is passed into avfilter_graph_create_filter(), occurs error in the following avfilter_graph_config() :

  [auto-inserted scaler 0 @ 32f77600] w:iw h:ih flags:'bilinear' interl:0
  
  [Parsed_null_0 @ 2e9d79a0] auto-inserting filter 'auto-inserted scaler 0' between the filter 'in' and the filter 'Parsed_null_0'
  
  [swscaler @ 3331bfe0] deprecated pixel format used, make sure you did set range correctly
  
  [auto-inserted scaler 0 @ 32f77600] w:1920 h:1080 fmt:yuvj422p sar:1/1 -> w:1920 h:1080 fmt:yuv420p sar:1/1 flags:0x2
  
  [libmp3lame @ 2e90a360] Channel layout not specified
  
  [in @ 3866e8a0] tb:1/44100 samplefmt:s16 samplerate:44100 chlayout:0x3
  
  [Parsed_anull_0 @ 330e8820] auto-inserting filter 'auto-inserted resampler 0' between the filter 'in' and the filter 'Parsed_anull_0'
  
  [auto-inserted resampler 0 @ 330e8dc0] [SWR @ 3809b620] Rematrix is needed between stereo and 0 channels but there is not enough information to do it
  
  [auto-inserted resampler 0 @ 330e8dc0] Failed to configure output pad on auto-inserted resampler 0

  I’ve googled for days but didn’t find any clue for it. Doesn’t what anull do is only "Pass the audio source unchanged to the output", why libav want to resample stereo to 0 channel ? What’s going wrong ?