Jakob

Integrity is priceless.
Communication Architect, Designer, Consultant. I use likes as bookmarks. @[email protected] |

[Jakob @jochmann]

@Klingebeil Yes and(!) maybe the whole mental model around intellectual property is fundamentally broken and artists need different modes of recompense that would free them and their work.

[Jakob @jochmann]

@Vjeux There are workarounds to route audio into the mic input (virtualcable, soundflower etc) https://t.co/ShzPGUeonk but that's probably not a dependancy that you want your users to install manually? Or at least I have not thought through, if there is a clever alternative.

[Jakob @jochmann]

@Vjeux 👍

[Jakob @jochmann]

@Vjeux Anytime. Is there a repo I could follow (and borrow from) if you happen to develop solutions that we might build on? Obviously I'm mostly interested in making a privacy respecting solution (like with local/self-hosted Whisper) performant. Thx!

[Jakob @jochmann]

@Vjeux You know you could have just, well, googled it:

https://t.co/DWRp1bdnDu

[Jakob @jochmann]

@Vjeux There's the google-API, no need for weird workarounds, if you are using chrome, btw.

We currently use it in https://t.co/xuKZTIaMh4 but would like to switch to viable alternative that respects data sovereignty. Only problem is making cost out of pocket bearable for non-profit.

[Jakob @jochmann]

@Vjeux If we're even going the route of processing in the cloud, we're looking to push the audio via websocket to a dedicated processing server and grab the text output from there. Cost/benefit is tricky.

Helpful starting points: https://t.co/aIsEm4wXVx and https://t.co/CJWnNzZyDu

[Jakob @jochmann]

@Vjeux To be usable at >95% accuracy you'll need to process the larger models, preferably via Cuda (Nvidia pipe) on GPU. We're testing a live captions accessibility tool & are probably going to have to run it on metal of our own.

If time is no concern, a typical server setup suffices.

[Jakob @jochmann]

@ellen_sch @ChElm @nadia_z @xhgMattia @CharlieBeckett I'm trying to find out, if "simple language" is part of the GPT-3 training set/output options (in various languages).

We're planning on developing a pipeline with whisper-AI for automatic captioning at events. Extending with automatic "simple language" excerpts looks promising.

[Jakob @jochmann]

@YoloLivTech I've been trying to send info both via the popup box for black friday promotions and your contact form and get a "server error". How can I contact the promo team to inquire about a deal for a special accessibility project with a local non-profit organization? Thanks

[Alice Evans @_alice_evans]

Sorry to side with our coming overlords, but “AI Soviet Afrofuturism” is fantastic!!

Cc @economeager @ATabarrok

[Jakob @jochmann]

@SachaGreif Katamari Damacy LARP meeting at Hachiko?

[Jakob @jochmann]

@maartenzam And there was this hint on Mastodon (shared by @sharoz https://t.co/6SNYGA5Ed9) using google docs I believe and this code snippet: https://t.co/dOZk6U4LBO

[Jakob @jochmann]

@maartenzam Same for me :D - it is on the road map according to the issues I read. You can always cook your own parser, if you have time and know-how. I don't have either at the moment so I'll be patient with this one or look for other tools.

[Jakob @jochmann]

Looks like I requested my Twitter archive just in time. I finally managed to download. Parsing with https://t.co/pgsIC1AOiz turns it into a nice human readable format.

While we can't replicate experiences, I do enjoy the pioneers' vibe and emerging culture at Mastodon.

[Jakob @jochmann]

@eagereyes @sharoz @richarddmorey It took a few days and now the download fails. Will retry later.

[Jakob @jochmann]

@kamilkazani privilege is a hell of a drug

[Jakob @jochmann]

@cassiecodes @smashingmag turning coffee into code?

my friend, do you know how much contemplative time I spend handgrinding, prepping and pulling my espresso shots on a manual lever machine?

cosycore all day.

[Jakob @jochmann]

So now that I've activated sharing, let's see what this does hm? #tm