Speech-to-Text.cloud

@SpeechToTextCloud@techhub.social
11 Followers
26 Following
149 Posts

Dedicated to high-quality #SpeechToText #transcription #STT

Personally interested in #Python #OpenAI #WebDevelopment #FireFox #fedi22

Websitehttps://www.speech-to-text.cloud/
Bloghttps://www.speech-to-text.cloud/blog/
Speech-to-Text.cloudJun 6
Martin HollandJun 6

Was lange währt ...

Wir haben jetzt eine eigene #Peertube-Instanz und beim Namen haben wir uns dann doch nicht ausgetobt. 🙈@ct_3003 ist schon dort, weitere Accounts sollen folgen.

Bitte schön folgen, wir wollen ja hier im Haus auch zeigen, dass das Interesse da ist. Und @keno3003 freut sich dann auch.

#Fediverse #Videos #ct3003

Speech-to-Text.cloudJun 6

Exciting updates at speech-to-text.cloud!

We've just added speaker identification, so you can easily label and track who's speaking in your transcripts.

Plus, our Summarize and Translate tasks just got a 30% speed boost!

Try out our improved platform and discover a smarter way to work with audio and video files.

https://www.speech-to-text.cloud/

#speechtotext #transcription #productivity"

Speech to Text Conversion

Upload an .mp3 and download your text transcript immediately. No subscription and no account required. Super-easy Speech to Text conversion.

Online Speech to Text Cloud
Speech-to-Text.cloudFeb 21
daniel:// stenberg://Feb 21
Sometimes people think #curl is a simple little HTTP tool, while in reality there's a whole internet transfer machine in there supporting 28 protocols.
Speech-to-Text.cloudFeb 21
Jeff TriplettFeb 21

🍹 Nothing like helping a client out with a server issue while they are away on vacation to get the blood flowing on a Friday afternoon.

🥳 Periodic reminder to contact my company, REVSYS, if you want to take a vacation without worrying about your servers while you are gone. https://www.revsys.com/contact/

We can babysit your apps while you are on vacation-as-a-service.

Contact REVSYS

REVSYS
Speech-to-Text.cloudFeb 1
Fabian N. T.May 24, 2024

aprxc — A command-line tool to estimate the number of distinct elements in a file/stream using Chakraborty/Vinodchandran/Meel approximation algorithm¹.

▸ Easier to remember & faster than `sort | uniq | wc -l`.
▸ Bound memory usage
▸ Exact until 80k unique values, 0.4–1% deviation beyond
▸ v2 with less BS in the README

Useful? You decide! Try out:
▸ uvx aprxc -h
▸ pipx run aprxc -h

https://codeberg.org/fa81/aprxc
https://github.com/hellp/aprxc

¹ https://arxiv.org/pdf/2301.10191#section.2

#math #ComputerScience #Python #CLI

aprxc

A command line tool / Python class to approximate the number of distinct values in a stream of elements using the (simple) Chakraborty/Vinodchandran/Meel algorithm (https://arxiv.org/pdf/2301.10191#section.2).

Codeberg.org
Speech-to-Text.cloudJan 31

The surprising way to save memory with BytesIO

https://simonwillison.net/2025/Jan/31/save-memory-with-bytesio/#atom-everything

#python

The surprising way to save memory with BytesIO

Itamar Turner-Trauring explains that if you have a `BytesIO` object in Python calling `.read()` on it will create a full copy of that object, doubling the amount of memory used …

Speech-to-Text.cloudJan 1
Simon WillisonDec 31

Here's my end-of-year review of things we learned about LLMs in 2024 - we learned a LOT of things https://simonwillison.net/2024/Dec/31/llms-in-2024/

Table of contents:

Things we learned about LLMs in 2024

A lot has happened in the world of Large Language Models over the course of 2024. Here’s a review of things we figured out about the field in the past …

Simon Willison’s Weblog
Speech-to-Text.cloudDec 31
Python HubDec 29, 2024

Uv, a fast Python package and project manager

https://docs.astral.sh/uv/

uv

uv is an extremely fast Python package and project manager, written in Rust.

Speech-to-Text.cloudDec 31
CassandrichMay 23, 2024

OMG! TIL there's a way to add back the long-since-removed capability to add your own custom search engines to Firefox.

In about:config, add a new boolean key browser.urlbar.update2.engineAliasRefresh = true.

Then the "Add" button is available in search preferences (put %s where the query term goes).

Source: https://superuser.com/questions/7327/how-to-add-a-custom-search-engine-to-firefox/1756774#1756774

How to add a custom search engine to Firefox?

Is there a way I can add a custom search URL to the Firefox search bar? e.g. I'd like to provide a URL such as https://blahblah.com?search=%s, where Firefox replaces the %s with the content of the ...

Super User
Speech-to-Text.cloudDec 26, 2024
LauchmelderDec 26, 2024
Ich teste seit einigen Tagen selfhosted #mailcow via #docker und bin bisher positiv überrascht. Installation und Setup sind sehr einfach und bisher läuft es geschmeidig und zuverlässig. Wie sind eure Erfahrungen dazu? Hat vielleicht noch jemanden einen guten Tipp dazu auf Lager?
×
Simon WillisonDec 31

Here's my end-of-year review of things we learned about LLMs in 2024 - we learned a LOT of things https://simonwillison.net/2024/Dec/31/llms-in-2024/

Table of contents:

Show threadNicole ParsonsDec 31

@simon

I wish there was more journalism on who funds these AI boondoggles & why.

We know a little about the motives of the bigger players but not the smaller ones.

Saudi Arabia believes it's buying capabilities that will shore up its regime.

US tech oligarchs believes it's buying capabilities for permanent one-party rule.

#KochNetwork believes it's buying artificially inflated energy usage to prevent a fossil fuel phase out.

Zuckerberg believes he's buying a method to avoid paying...
1/2

Show threadNicole ParsonsDec 31

2/2

... for content creation, media, & news.

They all think they're buying tax evasion & government subsidies.

They all believe AI will be an effective wage suppression scheme and it'll thwart unions & rising wage demands.

https://arstechnica.com/information-technology/2024/09/omnipresent-ai-cameras-will-ensure-good-behavior-says-larry-ellison/

https://www.washingtonpost.com/politics/2022/05/20/larry-ellison-oracle-trump-election-challenges/

https://www.bloomberg.com/news/articles/2024-11-06/saudis-plan-100-billion-ai-powerhouse-to-rival-uae-s-tech-hub

https://www.washingtonpost.com/technology/2024/05/14/middle-east-ai-tech-companies-saudi-arabia-uae/

https://www.businessinsider.com/openai-investors-latest-funding-round-2024-10

https://www.nytimes.com/2024/09/25/us/politics/kushner-private-equity-saudi-arabia.html

https://www.bloomberg.com/news/articles/2018-04-06/google-thiel-stand-out-in-saudi-prince-s-silicon-valley-tour

https://www.forbes.com/sites/kellyphillipserb/2024/09/28/tax-breaks-the-artificial-intelligence-is-changing-the-game-edition/

https://www.bloomberg.com/news/articles/2024-10-28/hk-unveils-ai-policy-for-finance-floats-tax-breaks-for-crypto

https://www.yalejreg.com/bulletin/belaboring-the-algorithm-artificial-intelligence-and-labor-unions/

https://www.thestar.com/business/drivers-worry-uber-s-new-pricing-algorithm-will-hike-fares-for-riders-while-reducing-their/article_43001daa-8592-11ef-b757-3fab1576058b.html

Omnipresent AI cameras will ensure good behavior, says Larry Ellison

“We’re going to have supervision,” says billionaire Oracle co-founder Ellison.

Ars Technica
Show threadGrant_HDec 31
@simon you touch on rate of change, but I think this, along with the learning curve to get consistently good results filter out a lot of people. I have a full time job, and need tools that work the same way each time I pick them up. The current rate of change makes any time investment in learning the current one a very high risk.
Show threadGregory P. Smith (he/him)  🚲🦝 Dec 31
@grant_h @simon That feeling is normal anywhere on the steep slope of the S curve of change. 2025 is shaping up to be fun in this regard.
Show threadErik MoellerDec 31

@simon

Excellent summary. I would add:

- vast, increasing numbers of recreational chatbot users (multiple 10M user category apps now); also increasing adoption of chatbots explicitly intended to be therapeutic

- massive adoption for search (Perplexity now valued at $9B, in large part due to continuing user growth); development of deep research tools like Gemini's & Scite.AI

- more learning/exploration focused models and applications (LearnLM, Perplexity pages, etc.).

Show threadScott JensonDec 31
@eloquence @simon Thank you for these references! I'm reading "The Future" by Naomi Alderman and she has a bonkers interpretation of the biblical story of Lot from Genesis. I used scite.ai to give me a summary, from Ms. Alderman's POV and got an amazing answer, with references! This feels so powerful, I can "reach into" books and get a more nuanced understanding of what it going on.
Show threadalcinnzDec 31

@simon On the topic of LLM criticism I will remark: I have tried playing a better role here, but each time I tried engaging in reasonable debate with someone the topic quickly turns to counter-factuals & I regret having started.

The whole space is very polarized, painting everyone engaged in the topic onto one side of the fence or other. So I minimize my engagement.

Show threadiveylineDec 31
@simon Interesting but what's the point in human terms - to ultimately make us all irrelevant?
Show threadSimon WillisonDec 31
@iveyline I really hope not. I like LLMs that augment human abilities - that give us new tools. That's one of the reasons I'm unexcited about the idea of "AGI" - that sounds like a human-replacement play to me, which doesn't interest me at all.
Show threadiveylineJan 1
@simon As a positive tool - fine. As a technology that manipulates and misinforms us - clearly not. So far the experience has been rather mixed.
Show threadlainyJan 2
@simon @iveyline every day hundreds of thousands of AGIs are born, so I wouldn't worry too much.
Show threadTodd BattistelliJan 2

@simon @iveyline Thank you Simon!

My team was laid off as part of an automation strategy.

What resources do you recommend to help workers obtain jobs where LLMs are used by as augmentation in well paid and dignified employment?

Show threadSimon WillisonJan 2
@FromTheRostra @iveyline I could attempt to answer that question for software engineering, I don't think I'd have anything useful or credible to say for other fields
Show threadColby RussellJan 1

@simon: "Want to build a Claude Artifact that talks to an external API? You’d better understand CSP and CORS HTTP headers first."

You don't need to understand CSP if you're not using it—and those who don't understand CSP *shouldn't* be using it. (As it turns out, "those who don't understand CSP" is most people. This is fine; CSP is an atrociously irresponsible design. Ideally it would be withdrawn as a standard.)

Show threadColby RussellJan 1
@simon even if we never get a mea culpa from the creators, browser makers should have known better.
Show threadSimon WillisonJan 1
@colby CSP is the answer to the question of why Claude Artifacts can only load external code from CDNJS - but the only reason I know that is that I spent quite a while digging around in the Firefox DevTools trying to figure out how their sandbox worked!
Show threadColby RussellJan 1

@simon there's probably a well-known image macro format that's a good match for the scenario.

Tries to make website secure with CSP to block third parties' code.

Whitelists CDNJS.

(What I didn't say in my previous post was that even a bunch of people who think they understand CSP actually have pretty much no idea what they're doing with it, and this is Exhibt A.

Almost everywhere you look with CSP, it's cargo cult all the way down.)

Show threadSimon WillisonJan 1
@colby I think they chose cdnjs because it does at least have a review process before libraries are added to it, unlike unpkg which serves anything pushes to npm
Show threadSimon WillisonJan 1
@colby since the whole point of Artifacts is running untrusted user-created code my guess is the CSP is more to prevent malicious artifacts from making unwanted fetch() calls to random domains
Show threadColby RussellJan 1

@simon humility is one of my favorite attributes, and I like when people who come across a "gate erected across a road" make an effort to consider that it might've been the product of wisdom and expertise and not incompetence, but some things really do come down to a group of Rhesus monkeys in a room with a stepladder flying off the handle without knowing why.

<https://kosmos.social/@colby/105947325146280629>

Colby Russell (@colby@kosmos.social)

Having spent a lot of time and energy trying to synthesize a useful description and associated shorthand for a common pattern that I see a lot—roughly, the inverse of Chesterton's fence—I recently realized the similarity of this pattern to something we already have a term for: the Just-World Fallacy. @dredmorbius@toot.cat

kosmos.social
Show threadColby RussellJan 1
@simon blocking fetch to arbitrary domains doesn't require denying people the ability to load (unreviewed) code from NPM. They're two different parts of CSP functionality, and blocking NPM-sourced scripts is neither necessary nor sufficient to stop those fetches.
Show threadSimon WillisonJan 1
@colby how would you block fetch() to arbitrary domains without using CSP?
Show threadColby RussellJan 2

@simon that's not what I said. The part of CSP that controls whether you can execute scripts sourced from domain X can be different from the part that says which domains images can be sourced from which can be different from the part that says which domains you can fetch/XHR.

Even ignoring that, whether CDNJS/NPM are reviewed/unreviewed isn't going to give just anyone who publishes packages control of the domains in question (to set up an arbitrary endpoint).

Show threadColby RussellJan 2
@simon in a better world, the answer would be, "You run the script in a worker and whoever is spawning workers gets to control that ('can the worker perform this fetch or not') capability."
Show threadKyle R. ConwayJan 1

@simon Thanks for this great write-up.

The disconnect between what the general population thinks these tools can do is concerning. This past year I've seen leadership in the non-profit space start using these tools in ways that correspond with the hype but not the reality. I haven't seen it be disastrous yet, but it's been confusing, disheartening, and disappointing -- often leading to much additional work.

Your point on LLMs needing better criticism is also a *very* good one.

Thanks much!

Show threadSimon WillisonJan 1

@K_REY_C that's part of this I feel most detached from at the moment - I don't have a great feel for how "real" people are using this stuff, just the occasional anecdote (many of them concerning)

One of the things I'd love to see in 2025 is a LOT more study and writing and reporting on how actual end-users are engaging with all of this

Show threaddrs1969 (David Smith) 🇬🇧Jan 3
@simon Just wanted to say thanks for this excellent writeup of all your work over the year. One of the best articles on where we are with AI that I've read. Much appreciated!