Does this need to connect to openai or does it function fully independently? Its for offline use.
this post was submitted on 28 Aug 2023
293 points (98.3% liked)
Selfhosted
38768 readers
393 users here now
A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don't control.
Rules:
-
Be civil: we're here to support and learn from one another. Insults won't be tolerated. Flame wars are frowned upon.
-
No spam posting.
-
Posts have to be centered around self-hosting. There are other communities for discussing hardware or home computing. If it's not obvious why your post topic revolves around selfhosting, please include details to make it clear.
-
Don't duplicate the full text of your blog or github here. Just post the link for folks to click.
-
Submission headline should match the article title (don’t cherry-pick information from the title to fit your agenda).
-
No trolling.
Resources:
- awesome-selfhosted software
- awesome-sysadmin resources
- Self-Hosted Podcast from Jupiter Broadcasting
Any issues on the community? Report it using the report flag.
Questions? DM the mods!
founded 1 year ago
MODERATORS
No, it's completely independent, it does not rely on any third-party APIs or anything else. It can function entirely offline once the models have been downloaded.
The readme mentions "transcription time on CPU" so it's probably running locally
How does it compare to https://github.com/guillaumekln/faster-whisper?
I've been using Faster Whisper for a while locally, and its worked out better than raw whisper and benchmarks really well. Just curious if there are any reasons to switch.
Whishper uses faster-whisper in the backend.
Simply put, it is a complete UI for Faster-Whisper with extra features like transcription translation, edition, download options, etc...
Nice! Thanks.
how does whisper do transcribing technical documents. like for lawyers, doctors, engineers and what not? or speakers with heavy accents?
Whisper models have a very good WER (word error ratio) for languages like Spanish, English, French... if you use the english-only models it also improves. Check out this page on the docs:
https://whishper.net/reference/models/#languages-and-accuracy
Congratulations on the launch and thanks for making this open-source! Not sure if this supports searching through all transcriptions yet, but that’s what I’d find really helpful. E.g. search for a keyword in all podcast episodes.
That's a great idea! I'll attempt to implement that feature when I find some time to work on it.
Oh, awesome! Does it do speaker detection? That's been one of my main gripes with Whisper.
Unfortunately, not yet. Whisper per se is not able to do that. Currently, there are few viable solutions for integration, and I'm looking at this one, but all current solutions I know about need GPU for this.
VERY understandable, requiring a GPU would limit it's application and spread, i hope a good GPU-less solution is found eventually
Awesome will give this a try
Nice, congrats!
I am looking for open source live transcription software, does this offer that, or is it only file-based?
I saw your project on Codeberg before. Then it was whisper plus. Since whisper+ it did not work anymore for me. I uploaded a file and it did not start. The old whisper worked. Did not try it for months anymore with whisper plus.
Maybe I give it another try. Can I use bind mounts or are there special permissions? Anyway thanks for your work.
Whisper+ had some problems, that's why I rewrote everything. This new version should fix almost (maybe there are some bugs I haven't found) everything.
If you take a look at the docker-compose file, you'll see it is already using bind mounts. The only special permission needed is for the LibreTranslate models folder, which runs as non-root with user 1032.
Massive kudos. I had the need for something like this in the past and it would have been a blessing.surely it will be for somebody else
I've been looking for a tool to do this for YEARS, my god! Years!!! ❤️❤️
Would love to deploy this, but unfortunately I'm running server equipment that apparently doesn't support MongoDB 5 (Error message MongoDB 5.0+ requires a CPU with AVX support, and your current system does not appear to have that!). Tried deploying with both 4.4.18 and 4.4.6 and can't get it to work. If anybody has some recommendations, I'd appreciate hearing them!
Edit: Changed my proxmox environment processor to host, fixed my issue.
I'm glad you were able to solve the problem, I add the comment I made to another user with the same problem:
Didn't know about this problem. I'll try to add a MariaDB alternative database option soon.
Congrats, and thank you for releasing this!
Maybe there's a couple of personal projects I could use it for...
This is excellent timing for me. I was just taking a break from working on setting up whisper.cpp with a web front end to transcribe interviews. This is a much nicer package than I ever had a chance of pulling together. Nice work!
Just tried this out but couldn't get it to work until downgrading mongo to 4.4.6 because my NAS doesn't ha``ve AVX support. But then, mongo stays unhealthy. No idea why.
Didn't know about this problem. I'll try to add a MariaDB alternative database option soon to solve this.