frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Modern C++ – RAII

https://green7ea.github.io/modern/modern.html
33•green7ea•2h ago•33 comments

The radix 2^51 trick (2017)

https://www.chosenplaintext.ca/articles/radix-2-51-trick.html
227•blobcode•7h ago•35 comments

Radio Astronomy Software Defined Radio (Rasdr)

https://radio-astronomy.org/rasdr
18•zeristor•2h ago•4 comments

Bridged Indexes in OrioleDB: architecture, internals and everyday use?

https://www.orioledb.com/blog/orioledb-bridged-indexes
13•pella•52m ago•0 comments

Tokenization for language modeling: BPE vs. Unigram Language Modeling (2020)

https://ndingwall.github.io/blog/tokenization
13•phewlink•2h ago•0 comments

Atomics and Concurrency

https://redixhumayun.github.io/systems/2024/01/03/atomics-and-concurrency.html
17•LAC-Tech•2d ago•1 comments

Practical SDR: Getting started with software-defined radio

https://nostarch.com/practical-sdr
159•teleforce•9h ago•43 comments

Turn a Tesla into a mapping vehicle with Mapillary

https://blog.mapillary.com/update/2020/12/09/map-with-your-tesla.html
37•faebi•1d ago•14 comments

Triangle splatting: radiance fields represented by triangles

https://trianglesplatting.github.io/
90•ath92•7h ago•37 comments

WeatherStar 4000+: Weather Channel Simulator

https://weatherstar.netbymatt.com/
621•adam_gyroscope•19h ago•115 comments

FLUX.1 Kontext

https://bfl.ai/models/flux-kontext
394•minimaxir•17h ago•99 comments

Show HN: MCP Server SDK in Bash (~250 lines, zero runtime)

https://github.com/muthuishere/mcp-server-bash-sdk
74•muthuishere•6h ago•19 comments

What Happens When AI-Generated Lies Are More Compelling Than the Truth?

https://behavioralscientist.org/what-happens-when-ai-generated-lies-are-more-compelling-than-the-truth/
9•the-mitr•1h ago•1 comments

Printing metal on glass with lasers [video]

https://www.youtube.com/watch?v=J0NNO91WyXM
5•surprisetalk•2d ago•1 comments

OpenBAO (Vault open-source fork) Namespaces

https://openbao.org/blog/namespaces-announcement/
44•gslin•8h ago•19 comments

The atmospheric memory that feeds billions of people: Monsoon rainfall mechanism

https://phys.org/news/2025-05-atmospheric-memory-billions-people-monsoon.html
27•PaulHoule•2d ago•5 comments

Dr John C. Clark, a scientist who disarmed atomic bombs twice

https://daxe.substack.com/p/disarming-an-atomic-bomb-is-the-worst
96•vinnyglennon•2d ago•63 comments

Why do we get earworms?

https://theneuroscienceofeverydaylife.substack.com/p/mahna-mahna-do-doo-be-do-do-why-do
6•lentoutcry•2h ago•5 comments

Player Piano Rolls

https://omeka-s.library.illinois.edu/s/MPAL/page/player-piano-rolls-landing
46•brudgers•8h ago•30 comments

Show HN: I wrote a modern Command Line Handbook

https://commandline.stribny.name/
353•petr25102018•20h ago•91 comments

Smallest Possible Files

https://github.com/mathiasbynens/small
42•yread•2d ago•16 comments

Buttplug MCP

https://github.com/ConAcademy/buttplug-mcp
179•surrTurr•4h ago•96 comments

How to Do Ambitious Research in the Modern Era [video]

https://www.youtube.com/watch?v=w7DVlI_Ztq8
31•surprisetalk•6h ago•1 comments

Superauthenticity: Computer Game Aspect Ratios

https://datadrivengamer.blogspot.com/2025/05/superauthenticity-computer-game-aspect.html
15•msephton•3d ago•5 comments

Show HN: templUI – The UI Kit for templ (CLI-based, like shadcn/UI)

https://templui.io/
37•axadrn•7h ago•20 comments

Show HN: Donut Browser, a Browser Orchestrator

https://donutbrowser.com/
43•andrewzeno•7h ago•20 comments

Making C and Python Talk to Each Other

https://leetarxiv.substack.com/p/making-c-and-python-talk-to-each
121•muragekibicho•3d ago•75 comments

Why is everybody knitting chickens?

https://ironicsans.ghost.io/why-is-everybody-knitting-chickens/
139•mooreds•2d ago•104 comments

I'm starting a social club to solve the male loneliness epidemic

https://wave3.social
215•nswizzle31•11h ago•391 comments

Germany eyes 10% digital tax on global tech groups

https://www.ft.com/content/39d4678d-a7e1-4fce-b8d8-eb799cfed3e6
52•saubeidl•1h ago•51 comments
Open in hackernews

Show HN: Typed-FFmpeg 3.0–Typed Interface to FFmpeg and Visual Filter Editor

https://github.com/livingbio/typed-ffmpeg
324•lucemia51•1d ago
Hi HN,

I built typed-ffmpeg, a Python package that lets you build FFmpeg filter graphs with full type safety, autocomplete, and validation. It’s inspired by ffmpeg-python, but addresses some long-standing issues like lack of IDE support and fragile CLI strings.

What’s New in v3.0: • Source filter support (e.g. color, testsrc, etc.) • Input stream selection (e.g. [0:a], [1:v]) • A new interactive playground where you can: • Build filter graphs visually • Generate both FFmpeg CLI and typed Python code • Paste existing FFmpeg commands and reverse-parse them into graphs

Playground link: https://livingbio.github.io/typed-ffmpeg-playground/ (It’s open source and runs fully in-browser.)

The internal core also supports converting CLI → graph → typed Python code. This is useful for building educational tools, FFmpeg IDEs, or visual editors.

I’d love feedback, bug reports, or ideas for next steps. If you’ve ever struggled with FFmpeg’s CLI or tried to teach it, this might help.

Thanks! — David (maintainer)

Comments

TechDebtDevin•1d ago
Good work!
yurifury•1d ago
Great idea. Personally looking forward to a typescript version of this.
CyberDildonics•1d ago
Typescript isn't mentioned anywhere.
yurifury•1d ago
Yes.
CyberDildonics•18h ago
What are you answering yes to?
internetter•18h ago
The fact that this particular implementation doesn't involve typescript has nothing to do with the fact that OP wishes for a typescript implementation. Your original response was a statement of fact that the poster already knew, so they said "yes" to affirm their knowledge
CyberDildonics•17h ago
No.
matt-attack•1d ago
I agree would love to see a typescript version.
teaearlgraycold•1d ago
God’s native tongue
barake•11h ago
Prompted the Jules preview to see what would happen. The implementation is pretty naive - there are much easier to read approaches I can think of. So not _awful_ considering the very short prompt I used.

Changes are on this branch for the curious:

https://github.com/matt-hensley/typed-ffmpeg/tree/feature/mu...

mikelinsi•1d ago
Cool!
BGZq7•1d ago
This is interesting, and it's good to see something that's actively developed, but it seems to have some of the same issues as ffmpeg-python:

- It doesn't appear to have any way of specifying filters that do not have inputs, such as "color"

- There is no way to provide flags to Popen, e.g. to specify subprocess.CREATE_NO_WINDOW to avoid CMD windows popping up in a GUI app on Windows. This isn't a big deal for running ffmpeg itself, because you can just ffmpeg.compile() then run it manually, but that can't be done for ffprobe with ffmpeg.probe().

Edit: OK, figured out the source filter thing, ffmpeg.sources.color. Is there a way to use arbitrary source filters, like vfilter/afilter can do for regular ones?

pbmahol•1d ago
Unrelated, but my C++ solution: https://github.com/richardpl/lavfi-preview
chrisallick•6h ago
this is what i was looking for: real time visual. very nice.
pietz•1d ago
That looks awesome.
cb321•23h ago
It is underappreciated that every single command-line option parser/toolkit is its own full configuration language with individual tools being "programs/configs" in that language. The lexical similarity of the zillion dialects (mostly due to Unix shells doing the word splitting for the eventual argv to be interpreted) masks what is really a dizzying diversity that for whatever reasons people think is much more uniform than it really is.

For example, I've done experiments running every single program in /usr/bin with --help and -h. The number of failures to get any useful help are a huge percentage. (The normalization of said percentage naturally is idiosyncratic to the exact system I ran that on).

Anyway, adding types to a complex one like ffmpeg may help more people realize this as well as offering practical benefits. So, great job!

plussed_reader•21h ago
For the unitiated where was the conflict?
cb321•20h ago
I'm not sure I understand your question. Are you asking for a list of the hundreds of commands which did nothing useful when run as cmd -h or cmd --help? I didn't save that, and as mentioned, it'd probably be different on your own systems/with your own packages.

If you want to try this at home, you should maybe either have GOOD BACKUPS HANDY OR DO IT WITHIN A VIRTUAL MACHINE/CONTAINER (EDIT: and almost certainly NOT AS SUPERUSER) also be ready to kill processes that are hanging waiting on stdin or somesuch. You're likely to have a least a few.

WITH THIS WARNING BEING GIVEN, you could just:

    (for cmd in /usr/bin/*; do echo "CMD: $cmd"; "$cmd" -h; done) >/tmp/dash-h 2>&1
    (for cmd in /usr/bin/*; do echo "CMD: $cmd"; "$cmd" --help; done) >/tmp/dashdash-help 2>&1
Of course, you could try /bin if that's any different (for me it's a symlink to the same dir as /usr/bin these days). You could also do single dash help.

If you want to accumulate some stats yourself, then you'll probably want to postprocess those output files.. So, you might also adapt the embedded echos to make it easy for whatever you like to do for that. Or, alternatively, you could re-direct each output to a per-command file with a little ${cmd##} massaging.

EDIT: and if you're asking what the CL syntax conflict was, well, I only meant to refer to "How to get any help at all - command with no args, with -h, -help, --help, -?, etc." as that is kind of the "very first question" on a user's mind. There are other syntactic conflicts (combining bool flags, option-value separation, unique prefix matching, etc., etc.).

aidenn0•19h ago
I know it's not the point you were trying to make but I find "man foo" to be more reliable and useful than "foo --help" or "foo -h"
cb321•17h ago
Fair enough. Finding documentation on the language is often step 1. :-) There are also giant POSIX PDFs around these days.

Also, I should have been more clear that while the CL parser toolkits are basically CfLangs/PLangs (maybe without loops, etc.), every command doing its own hand-rolled argv parser is basically its own totally unique language, probably doing only a subset of the 6..10 common conventions. It's probably not a great language, just something serviceable.

Basically, getopt wasn't good enough or easy enough or taught well enough, or done "soon enough". By the time it had come along, even within Bell Labs, they had probably grandfathered in old commands like `dd` and `find` (because "hey, I have a whopping 5 scripts that work with the old way!") with all their many & varied bespoke syntaxes.

Anyway, sadly, we only get our one try at "history" and many things are "sticky".

aidenn0•17h ago
It was even worse on DOS; wildcard expansion happened in the program, not in the shell, so you didn't know if wildcard expansion was supported for an individual command!
sedatk•12h ago
You still don't know if a Unix command would support processing multiple files even if the shell would expand wildcards. The tool might still process a single file and quit.
cb321•2h ago
Another issue is that, around the same time getopt was appearing, the Bourne family of shells added `key1=val1 key2=val2 mycmd` syntax, accessed more simply with either `getenv("key1")`, etc. in a C program or simply `${key1-default1}` in a script, or in python `import os; E=os.environ.get; E("key1","default1")`. This is very lightweight on the mycmd impl as it can just start using the parameter wherever, possibly with string -> native value conversion, if needed.

So, though "shell specific" (right at a time when that started to mean anything), BOTH filename pattern matching aidenn0 mentions AND "long options of a sort" were built-in at the shell level on Unix. E.g., `help= mycmd` instead of `mycmd --help`. The inheritance of environment across fork/exec made these propagate which is both blessing (propagates through wrappers, so need not capture/repeat) and curse (have to worry about namespacing/re-use).

Basically, various timing effects had outsized influence - Unix & shell diversity and right around the time GNU introduced `mycmd --key[= ]value` combined with portability agendas, and possibly syntax/properties led to long options winning. The value proposition was "Use 'cmd --key=value', not 'key=value cmd' and users with any shell will have easy syntax". I dunno..It's also hard to attribute winning.

Today, for good or ill, most people seem to literally be unaware of `key=val mycmd` calling syntax even though for bash/POSIX "beat" the C shell/most things decades ago, around when bash added the better !! history and command-line editing that made tcsh popular because mistakes happen/etc. I believe the fish shell also dropped this syntax, though.

ape4•23h ago
It seems that there should be a machine-readable description of the ffmpeg command line to generate this in multiple languages.
dejobaan•22h ago
The visual tool seems especially fantastic. FFMPEG seems an example of where modular/visual programming could really help, since I don't know all the bits FFMPEG offers. Minor UX note: I expected DEL to destroy nodes/edges (Win11/Chrome). But all in all: awesome!
userbinator•8h ago
It took this long for someone to finally come up with something resembling MS' DirectShow GraphEdit.
mertleee•18h ago
This is cool, but does it support piping frames between entire commands? In my opinion that's when you start to unlock the most interesting forms of FFMpeg flows.
pdyc•17h ago
this is great! let me plug my tool as well, in case you want to edit videos visually you can also use this tool https://newbeelearn.com/tools/videoeditor/ it generates ffmpeg commands.
Daiz•17h ago
If you're going to do script-based video processing in Python, I would highly recommend just going straight for Vapoursynth[1] instead. It's built for the purpose from the ground up, is actively maintained with a decent commmunity and tooling, and isn't tied to ffmpeg's CLI.

[1] https://www.vapoursynth.com/