FFmpeg 8.0 integriert Whisper-AI
https://linuxnews.de/ffmpeg-8-0-integriert-whisper-ai/ #ffmpeg #ai #ki #whisper #opensource #asr

some unsolicited knowledge sharing:
Here's a command you can use to convert H-264/AC3 MKV files to a browser-supported H-264/FLAC MP4 without any video transcoding (so no loss in quality). Audio streams are transcoded, but it's not really noticeable unless you have extremely high-fidelity inputs.

ffmpeg -i input.mkv -map 0 -c copy -c:a flac output.mp4

智慧眼鏡的即時隱私保護
➤ 讓您的智慧眼鏡應用程式在保護隱私的前提下，自由發揮創意。
✤ https://github.com/PrivacyIsAllYouNeed/protector
此專案提供了一套針對智慧眼鏡的即時隱私保護解決方案。透過在攝影機鏡頭與應用程式之間建立一個過濾器，該系統能自動處理即時影像，模糊或遮蔽未經同意拍攝的個人臉部，並可偵測語音同意，確保應用程式符合隱私規範。此方案強調離線運作，不依賴雲端，並支援多種輸出格式及友善的開發者介面。
+ 這項技術太棒了！終於可以安心使用智慧眼鏡來記錄生活或進行 AI 應用了。
+ 離線運作的設計非常吸引人，不用擔心資料外洩。期待更多應用場景的開發。
#智慧眼鏡 #隱私保護 #AI #FFmpeg #OpenCV #離線處理

FFmpeg 8.0 adds Whisper support: https://code.ffmpeg.org/FFmpeg/FFmpeg/commit/13ce36fef98a3f4e6d8360c24d6b8434cbb8869b
#linux #update #release #foss #ffmpeg #whisper

I don't watch porn

Whisper (transcription) filter in ffmpeg coming soon

https://github.com/FFmpeg/FFmpeg/commit/13ce36fef98a3f4e6d8360c24d6b8434cbb8869b

Hostias. Todo el finde liándola con #ffmpeg para concatenar varios archivos .VOB (vídeos de un DVD) en un único archivo con problemas porque justo en el corte se me corrompe la salida.

Al final la solución resulta que es... ATENCIÓN:

$ cat *.VOB > OUT.VOB

Funciona perfectamente y tarda menos de 10 segundos para un vídeo de dos horas. Me estoy despollando xDDDDD

We need a new open source license that states explicitly that the code may not be used in any situation where it will cause harm to people.

Where would Google be if they had #curl, #ffmpeg and a thousand other libs taken away from them?

Three lines of code just saved TuBoost.io users 2 hours of manual work.

Added automatic speaker detection to video cropping. Now when someone speaks, the camera follows them. No more static wide shots of empty chairs.

The magic happens in the FFmpeg filter chain. Computer vision meets video editing at 60fps.

Sometimes the smallest features have the biggest impact.

Kinda stuck on #FFMpeg.

I'm using libavcodec to decode a HEVC stream, but when I render it with #SDL3, the image is squished vertically, with bars above and below.

Strangely, I also have a hardware accelerated path working on RPi 5 (from their FFMpeg branch) using DMABufs and `eglCreateImage` and that is instead squished _horizontally_.

There's no crop or weird pixel aspect according to the AVFrame. Feels like I'm missing something simple.

Investigating why my server becomes unresponsive while running one specific ffmpeg command and find the process getting realtime (RT) priority. Huh?

I'm experimenting with av1 and turns out the av1 codec library requests RT priority! Wtf? Upgrading ffmpeg worked around this, but only because the RT priority request fails.

Very weird thing for a library to do imo.

For people who do amateur videos - I've found ffmpeg to do a magnificent job of reducing *.mov files.
For most of my short videos, the size reduction is 3-6x. For some larger ones, it can be >10x. Only once for a short video I've seen the size increase by a couple of %.

I use the following, and for my rare and simple needs there is no visible quality loss.

for m in *.MOV; do ffmpeg -i "$m" -vcodec libx265 -crf 28 ${m%MOV}mp4; done

Is there a (preferably FLOSS) tool I can use to turn .sup subtitle data into images, preferably batch.

I want to extract the subtitle images from videos to then run OCR on them. I know SubtitleEdit can run tesseract, but I find that tool clumsy and also tesseract output is really poor, so I'd like to experiment in doing the OCR with other things.

Get me the Marvel Comics timeline where we have MODOWOFCL: Mental Organism Designed Only for Working Out ffmpeg Command Lines ...

(yes, I've seen jwz's post about resize.pl. It's a lovely piece of code but it's particular about ffmpeg versions and is crashy for me)

(I've also seen that post about the AI that claims to be trained on ffmpeg command lines, thanks)

#ffmpeg is mighty and can probably solve your video encoding-related problem

Right now, it's re-encoding some unnecessarily high resolution videos I have into something much more storeable

e.g. 7 GB to 140 MB

https://github.com/jdriselvato/FFmpeg-For-Beginners-Ebook

Today would be a good day to donate to #FFmpeg. It’s the driving force behind so many things, and it surely deserves all respect.

https://ffmpeg.org/donations.html

I just had the idea that it’d be cool to create a FUSE filesystem that transcodes modern audio formats to MP3 on the fly, so I can play my music collection (mostly AAC and ALAC) in Winamp on one of my Windows 98 machines. Turns out somebody already did this:

https://github.com/nschlia/ffmpegfs

Wondering how well this works. Has anybody tried it?