Search results

Depends entirely on the bitrate used to encode the original audio. What matters most is the quality of the audio. With equal size, between mp3, m4a(aac) and opus, opus will sound closer to the original, followed by aac, followed by mp3. And if you encode the original audio to either of those 3...

The problem with comparing setting is that whisper isn't reliable. Even using the exact same settings, you'll get a different number of lines every time(never tested the same audio more than 4 times though but that's a lot already). No idea how it can sometimes fail and sometimes not with the...

It's a more modern audio compression format(m4a is aac) than mp3 and it's better at compressing than mp3 is. But opus beats both since it's even more modern and compresses even better so if you're going to convert it, which I wouldn't unless you're modifying the audio in some way, then better...

I'd say do a search for my user and limit the results to only this thread and with a bit of reading, you'll figure it out. Most of my posts lately are about helping people with making both the online and offline version of whisper work

No you don't. Purchasing computer units is just reserving a spot to use the remote computers to run whisper. It's completely free to run it on your own pc.

He last logged in the forum on october 11th 2021 so he hasn't been around for a while.

A troubled little sister who can't get out of diapers needs a big sister to take care of her. Diaper Club selection おむつの外せない困った妹には、お姉ちゃんのお世話が必要です おむつ倶楽部selection https://www.sanwapub.com/products/detail.php?product_id=21871 Studio: Abnormal Company Label: Diaper Club selection Director...

New links for SCN-036, SCN-057 and SCN-068.

Pacifier Prep School 70 おしゃぶり予備校70 https://www.knights-visual.com/works/furasupi/oshaburi_yobiko/kv-230/ Studio: Knights Visual Label: Furasupi Director: Hituki(ひつき) Code: KV-230 Runtime: 122 min Genre: Hetero, Gokkun Cen/Uncen: Censored Actresses: Momoka Kato(加藤ももか) Release Date: 2020/10/24...

Girl to girl, friend to friend. Two people who are not in a physical relationship take private selfies lesbian filming. 女子同士 友達同士。カラダの関係のない2人がプライベート自撮りレズ撮影。 https://bibian-av.com/works/detail/BBSS066 Studio: Bibian Label: - Director: Unknown Code: BBSS-066 Runtime: 230 min Genre: Lesbian...

Also available on something like ijavtorrent if you don't want horrible streaming quality.

It's an audio format, like mp3 or aac. There's many software that support it, I use either ffmpeg or opus tools but both are command line. Since I get asked all the time something like that here and I use mostly command line stuff since I automate these things which doesn't make for the best...

I know someone else is working on making an easy to use installer for it where you can just drag your files and it'll do its thing but no clue if or when they'll release that. M4a use aac as the codec so it has better compression than mp3 since it's more recent, but not as good as opus since...

The whole point is to avoid an extra mp3 conversion. And whatever input you give to whisper, it'll change it to 16k mono wav. The best result you see are just random and do not depend on the format of the input. You can feed it the exactsame audio file multiple times and you'll get a different...

Me I'd just use ffmpeg which does all that in 1 step but it's a command line software so not the easiest thing to use. Many software allow saving to wav so you just need one that give you enough control to do that. With ffmpeg, it would look like this: "ffmpeg.exe" -i "input.mp4" -vn -ac 1 -ar...

Mp3 is always a lossy conversion, regardless of the bitrate and so is opus. You won't hear the difference at such a high bitrate but who knows how it can affect a computer software. Wav, on the other side, is a mathematically lossless conversion meaning that no information is lost during the...

1. If you're going to do some processing on it before uploading, you should also make it mono, you're gonna save half the size and whisper is gonna do that anyway. Also, unless you have a slow upload or are in a hurry, it's better to convert it to wav instead to preserve as much quality as...

Best would be to extract/demux the original audio rather than convert it. There's many tools and tutorial about it on the net if you use the term demux and the extension of your video file. It also has been discussed many times in the last few pages here.

Can't help you with so little information. When you first use a model, it has to download it and it's a pretty big file so that's probably what it was working on for those several minutes you saw.

I purposefully included the address bar in the screenshot from the post I made above yours so people would know, but here it is anyway, it's the WhisperWithVAD link people keep posting here: https://colab.research.google.com/github/ANonEntity/WhisperWithVAD/blob/main/WhisperWithVAD.ipynb And...

In case there were any doubts that something other than 16kHz worked with the online version, I did a test just now with a 3h52m video: I also included the resulting srt sub with no editing.

No, it does work, I tested this on the colab with a 3 hour 48kHz audio file and it worked perfectly. The whisper code will convert it to 16kHz internally, that won't change whether it's local or remote.

That's not going to do anything, it'll just get ignored and keep the original sampling rate, which for the vast majority of audio is 48k or 44.1k(don't think I've seen a different one than those 2 in a movie or music/audio file unless it's a weird file). It's almost certainly not going to be 16k...

There's no benefit to using wav instead of just copying the audio and wav files will be 10-11 times bigger on average so that means if you have to upload it, it'll take that much longer. I'd just do something like this instead after checking what the audio type is with mediainfo to set the...

From the screenshot, his video is a mp4 so I don't think mkvtoolnix works for that, at least it didn't before, only for mkv. Multiple people mentionned some so just pick one. i use mp4box with a separate gui but the new version complicates things so not the easiest to install. Edit: Googling...

The difference is not big, likely unnoticeable in most cases. but it could hurt recognition and almost certainly won't help if you convert instead of extract. A very expensive nvidia gamer gpu(RTX 3080 in my case) is what most people would have that can work well enough for it, but just to do...

^ That is very likely the solution, uploading an audio file(that you tested to make sure is working first). What that ffmpeg error means is the input file isn't detected as audio basically so either the uploaded file got corrupted, was corrupt to begin with or the colab version can't handle...

:kecewa: You're making me feel bad now since all I had to do the first time I installed it was to copy paste the command and run it, lol.

What I'm saying is that you doing that did nothing, you must have done something differently after installing it that fixed it(or your audio file was corrupt before). It's true ffmpeg is required, but only on the pc whisper is running on.

That doesn't make any sense, it's doing the processing entirely off of your computer once you upload the file so installing anything locally wouldn't help in any way. What that error means is that ffmpeg had an error while running, not that it's missing. More specifically, it's unable to create...

You gotta be careful since extracting(aka demuxing or demultiplexing) audio isn't the same as converting(aka re-encoding) audio to a specific format. Both will give you an audio file on its own as a result(so you can technically say re-encoding it is extracting it too) so the difference might...

絶対領域 むっちり太もも制服チラリズム 生脚アイドルの究極挑発 小悪魔ニーハイ美少女 https://s1s1s1.com/works/detail/SSNI618 Studio: S1 Label: - Director: Bungo Maeda(前田文豪) Code: SSNI-618 Runtime: 210 min Genre: Hetero Cen/Uncen: Censored Actresses: Yua Mikami(三上悠亜) Release Date: 2019/11/19 Container: MKV Video codec: x265...

That's the same error I got when I uploaded a corrupted audio file. I suspect the issue is that you're uploading video files instead of only the audio file. whisper can handle those fine but considering the colab has extra restrictions, it might not be able to use the extracted audio because...

There are ways to get around those restrictions if you get creative enough in the way you ask, like telling it to tell you a story about that or telling it to act as if it didn't have that restriction, stuff like that.

They all do, that's 2022 stuff and older.

The colab thing is an extra interface on top of whisper itself so it adds more chances for things to go wrong. But yeah, the randomness of the result is very off-putting to me even though when it works, it works fairly well but it does try to be too smart for its own good. We'll see if the next...

If you scroll down, where you were seeing more information about the error before, it'll tell you if it's doing anything. But since it says it's queued, that likely means the gpu is unavailable or something like that and it's waiting for a free spot to do its thing. If it completes, it'll open...

It is a very long filename and depending on how the system on the other side is configured, some characters ca be misinterpreted which can cause issues like you experienced. That's why you should always give as much information as possible when you're trying to get help, makes finding potential...

The GTX 1650 has cuda v7.5(same as the RTX 2070) capabilities which is likely supported by either version pytorch has to offer. Telling us the error you're getting is the first step to getting support to fix the problem, not much we can do without that information. But as mei2 said, the 4GB of...

Since MrKid didn't mention this method, here's exactly how I did the drag and drop method, for those having difficulties, if you want to make sure the issue is something else. 0. DO NOT run the Mount Google Drive step. 1. Click the folder icon to open the Files column on the left of the...

I haven't analyzed any of the subs I created with it in much detail, just messing around a bit with it from time to time so couldn't tell you how often it does it. I just tested to see if it supported dts audio from a bluray and the very first line had that issue. If you read the limitations...

Even direct english transcriptions aren't perfect with whisper, it can invent stuff it didn't hear but makes sense in the sentence. A funny one I encountered was before a song the guy said 1,2,1,2,3,4 to start it, but whisper continued counting until 13.

I just dragged it there. Could be ad or script block extensions causing upload issues or a browser that doesn't support the upload procedure so trying a different one might help, i used brave which uses the chrome engine. Could also be you're not waiting for the file to finish uploading, the...

You probably didn't copy the path properly. If you upload to the colab, it should be like this: Click the 3 dots you see when you hover over the file, copy path and then paste in the input. Doesn't seem very happy with an opus file for some reason(ffmpeg 100% support it) or I did something...

Ok, so it's not running it on your own pc but it the colab thing? If so, your gpu is irrelevant, unless you followed what Epinwinrar posted to install the webUI locally, in which case do what I suggested in my previous post to test it. It tells you it failed for some reason but doesn't say why...

The RTX 2070 "only" has 8GB of VRAM so the large model wouldn't work with it since it requires 10, but any other model will be ok and the cuda version should be fine although I can't find out the exact requirement for whisper. I have no clue how the web based version works, but it likely hides...

Should handle pretty much any audio file type, pretty sure it install ffmpeg as a dependency to deal with that. Without an error message, it's impossible to say what the problem actually is, but a likely cause would be that you're using a model too big for your GPU VRAM or you don't have an...

Not much point to splitting it unless you have more than 1 gpu to work on the chunks, it'll just slow things down on the 1 gpu. But to do it manually, you'd use something like ffmpeg with a time range when you convert the audio to mp3, change the range to make multiple parts, process the parts...

I don't want to be mean since not everyone is good with computers, but whisper is actually very easy to install. What's most likely happening to most people who failed is they did a small mistake somewhere because they're not familiar how command line works or they needed administrator rights...

It should still work with lower quality audio but the result will not be as good. soloporhoy666 is already lowering the quality to mp3s instead of using something like opus which would give better quality for the same size and is apparently still getting good enough results to get impressed.