Whisper in Subtitle Edit cannot see the audio file

[loninappleton]

I've forgotten a lot about how to get Whisper going in Subtitle Edit. File types mismatch maybe.

I have an MP4
I extracted the audio in Goldwave and saved as WAV

In Subtitle Edit I opened the Whisper dialog from the Video tab and dragged the WAV to
ADD. It's in there so no need to press ADD again.

then go to the bottom and press Generate.

At that point it does it's extraction briefly and then displays the message in the
screen shot. I have not seen this message and don't really know what it's about.
A blank SRT results from it.

[KarMa]

I find that blank SRT files result from either, not enough RAM being free during the process (ideally 5-10GB free or more), or your main boot drive not having enough room to store the WAV.

[loninappleton]

Thanks for answering. Yes I see that the small SSD I'm using is getting full. I'll retry when I clean out some work files.

[loninappleton]

I looked at the RAM on here and it's 8 Gb. Looks like that's the end of it.

[VoodooFX]

Press on "Engine" and select Faster-Whisper-XXL, your selected model would need only 1-2GB RAM.

[loninappleton]

Thanks for the tip. I'll see how it goes. For Engine, that is download the model?

[loninappleton]

Press on "Engine" and select Faster-Whisper-XXL, your selected model would need only 1-2GB RAM.

I found that in the engine area where CCP etc the default. Which model is best to pair with this that needs less memory?
And is the WAV still the best or only file type to use on this?

I remember other attempts to use Whisper where it may start out ok and then sort of get lost on a long translate. I have a 9 minute section to test when it comes to that.

[VoodooFX]

Which model is best to pair with this that needs less memory?

If YOUR best = less memory, then use tiny model.

And is the WAV still the best or only file type to use on this?

It was never "the best", use the original files, that means - don't touch anything.

I remember other attempts to use Whisper where it may start out ok and then sort of get lost on a long translate. I have a 9 minute section to test when it comes to that.

Use Faster-Whisper instead of Whisper.

[loninappleton]

I extracted a WAV as the largest and so most detailed or slow and easy to work with. But here is the MediaInfo of the original (below)
So extract to AAC-LC? Some of these details I'm just seeing for the first time. I see another user is on here so I'd want to put everything together from all suggestions.

MediaInfo
General
Complete name :[edit]******************* - Measure for Measure - Intermission Featurette (2004) - [720].mp4
Format : MPEG-4
Format profile : Base Media / Version 2
Codec ID : mp42 (mp41/mp42/isom)
File size : 587 MiB
Duration : 19 min 52 s
Overall bit rate mode : Variable
Overall bit rate : 4 130 kb/s
Frame rate : 29.970 FPS
Movie name : Featurette (Shakespeare's Globe, 2005)
Movie_More : Shakespeare's Globe, 2005
Keywords : iMovie
Encoded date : 2020-05-19 04:00:45 UTC
Tagged date : 2020-05-19 04:06:59 UTC
Cover type : Thumbnail

Video
ID : 2
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L3.1
Format settings : CABAC / 2 Ref Frames
Format settings, CABAC : Yes
Format settings, Reference frames : 2 frames
Format settings, GOP : M=1, N=30
Codec ID : avc1
Codec ID/Info : Advanced Video Coding
Duration : 19 min 52 s
Bit rate mode : Variable
Bit rate : 4 001 kb/s
Maximum bit rate : 768 kb/s
Width : 1 280 pixels
Height : 720 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 29.970 (30000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.145
Stream size : 569 MiB (97%)
Title : Core Media Video
Encoded date : 2020-05-19 04:00:45 UTC
Tagged date : 2020-05-19 04:06:59 UTC
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709
Codec configuration box : avcC

Audio
ID : 1
Format : AAC LC
Format/Info : Advanced Audio Codec Low Complexity
Codec ID : mp4a-40-2
Duration : 19 min 52 s
Source duration : 19 min 52 s
Bit rate mode : Constant
Bit rate : 128 kb/s
Channel(s) : 2 channels
Channel layout : L R
Sampling rate : 48.0 kHz
Frame rate : 46.875 FPS (1024 SPF)
Compression mode : Lossy
Stream size : 17.9 MiB (3%)
Source stream size : 17.9 MiB (3%)
Title : Core Media Audio
Language : English
Encoded date : 2020-05-19 04:00:45 UTC
Tagged date : 2020-05-19 04:06:59 UTC

Image
Format : JPEG
Width : 1 280 pixels
Height : 720 pixels
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Compression mode : Lossy
Stream size : 108 KiB (0%)

[VoodooFX]

So extract to AAC-LC?

No, no need to extract anything.

[loninappleton]

On extraction, that's the way I've done it. Are you saying just go ADD and get the mp4 in this case and it does the extraction
to AAC?.

In the meantime I have that shorter excerpt (an interview) to play with. These jobs normally take a while so I've set up my backup PC with even less memory that can just run by itself. 4Gb memory on the ol' MSI 760 gm. If that's a problem even with tiny model I'll have to forget doing that.

[loninappleton]

The setup of Whisper with these tips of tiny model and the special Whisper Faster version (above) have gotten good results.
There are corrections to make but I have something I can work with now.

thanks to all who answered.