search

tmoroney / auto-subs

Generate subtitles using OpenAI Whisper in Davinci Resolve editing software.
https://tom-moroney.com/auto-subs/
MIT License
578 stars 34 forks source link

My experience of installing the plugin for Windows 10 #66

Open NanashiEltnum opened 2 months ago

NanashiEltnum commented 2 months ago

Hi. This is not a bug report, but information about the nuances of installing this plugin. I hope it helps you as much as it helped me. I'm a complete dummy at this so far, I figured it out on my own. I have Windows 10 and beta Davinchi vers (19.0b build 43)

  1. Do the installation as written in the plugin installation. The only thing I didn't use was Chocolatey or Scoop. But about that in step 3.

  2. As written here before me, the version of Python installed on your computer must be the same as the one used by Davinci. Otherwise you will get the error "No module named "stable-whisper"". How to check this is here https://github.com/tmoroney/auto-subs/issues/42 You MUST uninstall old versions of python!!! I had a situation where I had an old version, 10.x, and a new 12.x version. Windows registered the old one, which caused an error. I don't use anaconda or anything like that.

  3. How to find a path to install ffmpeg without Chocolatey or Scoop. If there are no install ffmpeg files, Davinci itself will tell you the correct installation paths in the console. For me it is: C:\Users\%USERNAME%\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib.

  4. If you get an error that fbgemm.dll is missing, first update VC_redist (Visual C++). If the error continues, then throw ibomp140.x86_64.dll to the path from the 3rd item.

I hope these tips helped someone.

dkiaulakis commented 2 months ago

Thanks. As well worked to install visual studio (python and C++ desktop options). So far with 19.6 studio DR version couldn't make appear titles on a list and on timeline. audio.srt file was created from timeline. Same screenshot's shared on discord as well: 2024-08-13 14_35_27-Resolve 2024-08-13 13_32_38- 2024-08-13 13_24_49-

P.S. transcription does not use "fully CPU and GPU recourses, I've installed cuda package from Nvidia. How to make process faster which could use more computing resources availbale on win10 machine?

YourAverageMo commented 2 months ago

Additional notes

i want to add a few things i struggled with when installing it today after finally getting it working.

1. No module named "stable-whisper" like you said make sure your using the same version of python DR is using but also the reason why i was getting the "No module named "stable-whisper" error was because i didnt run powershell in admin mode while installing all the packages so pip was installing them in a different location and not adding the proper PATHS. So i simply launched powershell in admin, uninstalled stable-whisper, and reinstalled.

2. fbgemm.dll is missing error if you get the fbgemm.dll is missing error what fixed it for me was installing microsoft visual studio > desktop dev with c++ and the only tick box that mattered was this: MSVC V143 - VS 2022 C++ x64/x86 build tools (latest)

3. Numpy... also the numpy thing was pretty common and easy to fix. Just going to add that here for others aswell. all you need to do is uninstall numpy and install 1.26.4 using: pip install numpy==1.26.4

reference: https://github.com/tmoroney/auto-subs/issues/49#issuecomment-2178534662

lucas-mandai commented 2 months ago

For me, what worked was downloading ibomp140.x86_64.dll and adding it to the path. Thanks, this issue helped me a lot.

viewtifultiger commented 1 month ago

Yeah, having an older python version was the problem for me. The autosubs window now appears whereas before I kept getting the "could not find stable_whisper" error in the console