Nothing to watch – Experimental gallery visualizing 50k film posters | Hacker News Reader

Nothing to watch – Experimental gallery visualizing 50k film posters

nothing-to-watch.port80.ch

140 pointsbj-rn76 comments

Somehow I swirled around for a minute or so and didn't see anything I would want to watch.

"This gallery features a collection of the 50,000 most popular* movies according to TMDB"

Yeah, there's the problem. I went to the TMDB site and tried sorting by popularity, rating … all I got was a whole bunch of recent films that look like all the reasons the film industry is crashing (you know, the kind of stuff you would see in a Red Box when they were still a thing).

If you really don't know what to watch you could do a lot worse than checking off films on this list (the 1001 Movies to See Before You Die):

https://1001films.fandom.com/wiki/The_List

(And you have already seen a number of them I expect.)

Elitist? Maybe. But if you can actually get through the list (I've been at it, casually, for several years now) you'll feel like you have acquired a film school degree.

I think it's fine to watch "8 ½" and hate it — at least you'll know that you hate it now. But it would be sad if you never saw "La Strada" because you thought all of Fellini's films were impenetrable.

"Marty", "Stella Dallas", "The Hustler" — just a few great American films that came to mind that I would have otherwise missed…

But if the latest Guardian of the Pirate's of the Marvel Universe movie are your cup of tea you probably have no problem finding something to watch anyway and can pass on "The List".

The worst thing, or irony, or something, IMHO, about Marvel isn't even how bad or good any of them are in isolation. It's that they are so similar so close in (release) time to each other.

I'm sure one could make a "top 5 Marvel of the 783 released so far" movies and these 5 are actually pretty darn good.

The joke is though that the list of "top 5" could probably be picked from a pool of 15 good ones and it wouldn't matter. The 15 are all pretty good.

But if you watched all 15 good ones you'd be bored out of your mind because it's the same formula.

And if you watched all the ones which are merely "ok" it's like watching the more boring variants of anime where very little happens over very many hours.

(Specific numbers just made up for the argument.)

That list has Avengers Infinity War and Endgame on it by the way.

The list has 'Watership Down' that I found very quickly and liked

That list has Avatar and La La Land on it. It’s definitely not elitist!

There’s some real gems on lists like this but some of the older stuff would definitely be a slog for most people and would probably need to be put into context to make sense.

Recently I’ve engaged in a practice of just hitting play on any old movie on Prime with at least an actor or director or even studio I’ve heard of. Essentially any film from the 40s to the 00s that had a budget and was remembered enough to make it to Prime. It’s been quite rewarding, all sorts of hidden gems out there even from the 70s through 90s. Even the mediocre stuff from then is often more nuanced and creative than a lot what you get today.

> would probably need to be put into context to make sense

You're right. The book ("1001 Movies to See Before You Die") gives that context. I was surprised though that, as I went through them — slogging even, as you say, at times — in time I started to develop perhaps a new way of looking at film. And then slowly I found I was "slogging" a lot less. I began even thinking a bit more fondly back on the films I had seen earlier on that had been more of struggle for me to watch.

> with at least an actor or director or even studio I’ve heard of

I think that's the gist: anything goes, and maybe start with 3 of anything!

...based on genre (Comedy): Tampopo (1985), Matchstick Men (2003), Force Majeure (2014)

...based on subgenre (Heist): The Killing (1956), Thief (1981), Sexy Beast (2001)

...based on director (Kelly Reichardt): Old Joy (2006), First Cow (2019), Showing Up (2022)

...based on production studio (Studio Ghibli): Only Yesterday (1991), Porco Rosso (1992), The Wind Rises (2013)

...based on location (Vienna, AT): The Third Man (1949), Bad Timing (1980), Before Sunrise (1995)

...based on interest (Desires): Deep End (1970), Exotica (1994), Ghost World (2001)

...based on language (French): L'Année dernière à Marienbad (1961), Un homme qui dort (1974), Portrait de la jeune fille en feu (2019)

...based on movement (Czech New Wave): The Sun in a Net (1963), Valerie and Her Week of Wonders (1970), Morgiana (1972)

etc.

> based on subgenre (Heist):

If you have not seen the OG "Topkapi (1964)" you missed a real gem.

If you're only interested in great movies, no others allowed, then the top 50,000 is bound to disappoint as there aren't that many great movies yet. Setting it to 10k this is what I got in my first 5 (well spread) clicks:

- Ikiru (1952): A fantastic movie, also in the above list

- A Clockwork Orange (1971): Alright, nobody is going to believe these are random if I keep pulling these...

- Tracks (2013): Haven't seen this one, doesn't seem particularly astounding but I'd probably watch it, especially if my wife was wanting to watch something we haven't seen together.

- Songbird (2020): This one looks a bit garbage to me, probably wouldn't watch.

- Osmosis Jones (2001): Watched a few times, maybe a bit of nostalgia but it has been a few years so I would throw this up. Not deserving of the top 1000 by any means... but also not Leprechaun 5: In the Hood (2001).

And that seems pretty reasonable. I'm only about 800 movies in (since I've started tracking), of those only ~400 are in the original 1001 Movies You Must See Before You Die list. To me watching straight through the various compendiums of the best movies is actually more dull than mixing in other things like the latest pop films or some stuff that wouldn't make many "top" lists but isn't necessarily bad. It's a bit like touring the best restaurants in a new area - if you do them all in a row then you just lose some of the feeling of variety. I doubt Tracks will leave an impact on my soul to the level of Ikiru, but I still want to watch things like that often.

Osmosis Jones has Bill Murray. Anything with Bill Murray is likely not too bad. And so on. There are many little walking paths like that. :)

Yeah, disappointed that, even in your anecdotal list above, three of the five are in the last dozen years.

Two of the five, with not a single movie from the same decade. Given the production rate of movies has not been constant either, is this really more than the statement "I only want to see old movies"?

I don;t think so — it's just that if you have 100-ish years to choose from, this doesn't seem like a very even spread. To be sure, I would feather off anything too early — the silent era seems only interesting to film historians that want to see where the modern ideas of film began (example: "Battleship Potemkin" (1925)).

But at the same time, Hollywood does seem to have fallen into the trap of mediocrity for some time now — probably since the 1980's to be honest (having only gotten worse over the ensuing decades). "La La Land" was popular in large part because it was something of a throwback to 1930's Hollywood. Tarantino's success seems to stem from his mining of gritty 1970's Hollywood.

So why not go back and watch "Gold Diggers of 1937" for the "La La Land" stuff, or "Straw Dogs" (1971) to see what Tarantino is making homage to.

Without putting any assumption about whether new movies or old movies are more often better, wouldn't one expect a skew somewhat following the skew of rates of movies produced per year?

Strange list. Some great films, some absolute shit and glaring omissions [1].

[1] for my personal taste

Love to hear your omissions.

And as for absolute shit? Yeah, there's Andy Warhol on there.

Hey now... Flesh for Frankenstein is a fun weird mess, at least.

Sharing some in my all-time top ratings which are not in the list (would be cool if other folks shared some of theirs as well):

- Everything Everywhere All at Once (2022): Newer than the cutoff date of the list.

- Your Name (2016): I'm not usually a fan of Anime but this newer film hit different.

- Sunshine (2007): This one I know I consider far better than most do, but it's one of my favorite variants on this kind of story for some reason. Also Cillian Murphy is my favorite actor in the modern era.

All 3 of these happen to be from after the original 2003 book from Steven Jay Schneider as well (though they aren't in the newer list either). I find he did a really good job of covering the classics, but I'm sure if I dug through my top 100 I'd hit something older I liked a lot that didn't make the cut. I'm just too lazy to :D

"One, Two, Three" (1961) is Cold War comedy gold.

"A Tree Grows in Brooklyn" (1945) is a real tear-jerker even 70 years later.

One, Two, Three was great! I hadn't heard of A Tree Grows in Brooklyn, added to the watch list for sure. Thanks

There’s Toy Story 1-4 on there and Star Wars Force awakens, but for popular films it doesn’t have the incredibles, no Star Trek (perhaps include ii), few/no bond films (Casino Royale?), few Miyazakis (Nausicaa, Mononoke), not many Tarantinos (hateful 8), not too many Burtons (Batman 2, Big Fish, Sleepy Hollow). Is there any Indiana Jones? Could include Starship Troopers.

Just a few:

Kin Dza Dza.

Tomorrow I will wake up and scald myself with tea.

Big trouble in Little China.

My neighbour Totoro.

real elitists are up to their eyeballs in criterion collection backlog, or so I like to think.

I thank god for Criterion!

Criterion is the antidote to YouTube shorts and TikTok

I discovered I love just about every movie Bogart was in through Criterion

Snatch (2000) isn't on it? Blasphemous.

I'll add it – two minutes Turkish.

Lock, Stock, and Two Smoking Barrels is on it, which arguably is a better film.

I did not know of this list -- thanks! According to my Letterboxd, I have seen 757 of them already.

It seems that there are different "editions" of the list, where new films are entered and some are removed, to keep the count constant.

I appreciate the effort but, this in many ways showcases the design problems of today that put form over function: high score for the FX, low score for usability.

This is fricking awesome. It's the coolest thing I've seen on HN for a while. Love it!

N.B. I visit HN multiple times per day. I've been a member of HN since 2010. My last comment was made in 2018. I logged in for the first time in years just to make this comment :-)

It really feels like something out of "Web 4.0", like some random next-generation UX pattern (for better or for worse)

I agree, but then I spend a lot of time talking about and developing software for Apple Vision Pro so perhaps I see this in a different light to most. Feels like a great interface for random discovery and/or semi-random discovery within more tightly connected bubbles of taxonomy.

I don't get it. Good visualization takes complexity and brings clarity and focus to it. There's seemingly nothing connecting where things are placed or how moving in any direction will lead you somewhere that develops understanding of what's going on. It's just a massive grid with magnification and a pseudo voronoi effect.

Author here - thanks for the kind words!

Author here. The Cloudflare traffic led me here - apologies for all the bugs and crashes. This was semi-leaked. I only shared the link on a small subreddit yesterday, I'm aware of loads of bugs and it’s never been tested on anything beyond a ThinkPad and an iPhone.

To those criticizing the UI: agreed. To those criticizing the movies: also agreed :D

As some commenters have already pointed out, it’s more of a tech demo than anything.

https://opensearch.westloop.io/poster

We created this as a demo for OpenSearch on AWS (we are AWS partners) but I really enjoyed making it and I thought it turned out to be really fun.

Took the same type of dataset at the OP (I think we have 80,000 posters) but we used AWS’ multimodal embedding model to enable semantic search across the posters, as well as the ability to do a similarity search based on the posters.

Eventually we will include AWS’ personalize as part of the service to show off some of the movie recommendation engine features they developed.

Not commercialized there’s definitely some bugs, we built it for fun

Ohhh this is really cool!

A little bit better way to navigate back and forth on results would be amazing.

But loving it, nice demo.

where do I get the dataset?

I have no idea what is going on or what is supposed to be going on. I opened up the console to see error messages, but there were none. But I'm pretty sure it's not rendering the way it's supposed to.

It's right there in the title: there's nothing to watch.

(Not working on Firefox mobile, but my guess is hugged to death)

Tried different modes, but does not seem to work for me. Console just spams handlePointerOut hundreds of times.

Once in depth mode I saw movie posters but it zoomed out until they where gone.

I had a look, then heard the fan on my pc ramp up...

32.0 GB (27.8 GB usable)

AMD Ryzen 7 8845HS w/ Radeon 780M Graphics 3.80 GHz

I have 64Gb, a 9800x3d, and a 4070Ti super. It's still a little slow if I set it to the highest settings, but it's usable.

Very cool either way!

Make sure to share the browser too :)

Chrome in Windows 11

I think shows and movies used to feel more special back when media was harder to access. We had to wait for them to air on TV, and if we missed them, that was it. The scarcity made watching them more exciting and meaningful. Now that everything is instantly available, it all feels cheap and disposable. In my opinion, this is one of the reasons streaming services will eventually fail. If they want to survive, they need to find a way to bring back that sense of anticipation and urgency.

I love it. The UX is terrible but the visualization is very "outside the box". Experiments like this are important on the road to finding novel interfaces. Not everything has to look like a v0 shadcn app. Thank you for sharing!

“Nothing to watch” is actually well solved by LLMs. Just give them a list of your favourite films and ask for some less known of films which are in a similar vein.

An ocean of film posters is a cool visualization but doesn’t solve the problem well.

Anyone else remember Microsoft Silverlight? When it came out, they had a C# & Silverlight published demo that was a movie library viewer similar to this -- not as nice, but good for its time (circa 2007).

I had a boss who was bought into .NET to the point where he thought maybe the future was Silverlight (instead of JavaScript, which had obviously "won" at that point) and he wanted me to make Silverlight have the features that the desktop framework had at the time. Which I told him immediately that seemed impossible, but since he was so sure about it I would be happy to try (I had just been given a contract there). So I was decompiling .NET framework or WPF libraries or something (don't remember) and trying to get them to compile in Silverlight. I don't remember how long I spent on that task. Possibly over a week.

I had another job last year where they kept asking me to fine tune an LLM to answer questions about text, but I was only allowed to feed in the raw text of a single document but not generate an actual Q&A dataset. I kept telling them that was not how it worked but they kept insisting I continue for weeks. I think in his mind he just thought I was just doing it wrong over and over and he was secretly training me by paying me to practice or something.

I guess it's kind of sad how many jobs I've had that were a waste of time. I'm weirdly used to it so looking back it doesn't bother me that much somehow.

It does irritate me in the moment though. I have a client now that wants to make an agent system for a certain platform. The previous day we had just been discussing MCP servers and what they do. But the next day he seemed to think that each MCP server was going to have an entire full custom UI and agent loop in it per agent instead of just providing tool commands. So MCP server became an agent. I think I explained how it normally worked and he seemed to maybe get it, but then I was trying to ask how we were going to allow people to use all of the existing MCP servers in their agents so they could combine them and the non-response made it seem like he thought that might not be important. Did not get an answer so far. I remain ever hopeful for our upcoming meeting though.

Is this a sign of old age? Answers to something specific just become rambling walks down memory lane? It's interesting how it seems that I am generally well adjusted to how futile my life has been.

Anyway, I'm sure it's just a coincidence that Silverlight sent me down this tangent.

Really cool website! What inspired you to create it?

Just a heads-up: it's broken in Safari, because the app-*.js is served with Content-Encoding: zstd, which Safari doesn't support.

I'm guessing you mean Safari on iOS, because it works just fine on macOS.

No, I'm on macOS — Sonoma though. Interestingly, MDN lists no support for any version: https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/... — apparently the table is outdated.

Interesting, I'm on Sequoia with Safari version 18.5 (20621.2.5.11.8).

Yep, MacOS (Sequoia) and it works for me as well.

As someone with Tripophobia I really, really didn't like the "depth" option, but otherwise like the idea; seems like it could do with showing more of the posters and titles though as at one point about 10 flipped by in a second before I could read more than one or two of the titles and saw nothing but a blur for the posters.

Some time ago, I wanted to create a website that would recommend movies based on rating from people whom you have similar movie ratings with but I was unable to get the necessary data as existing websites guard it. But it would be nice to finally have a functioning movie recommendation engine one day...

how to go back to device selection? I clicked hi-end but its slow on MBP M4 Max 64Gb

There's a gear icon in the top right corner to change the settings.

Nice idea, horrible interface

When I don't know what do watch I simply check the list of Criterion Films and pick the first one that I recognize the title but haven't watched yet. So far it has worked very well for me.

Estimated my work laptop as "high end" and then Edge crashed.

Is this the new furmark?

When I saw "Your device" I assumed it is about device for watching, not about specs of the currently used device.

It should be communicated in a clearer way.

I found the labeling "potato" "mid-range" and "High-End" very clear.

Whoa, the first film poster it presented to me was for Koyaanisqatsi, a 1982 film I was unfamiliar with until I saw it in a Metafilter post just this morning[1]. Baader-Meinhof in action[2].

[1] https://www.metafilter.com/209750/Godfrey-Reggios-Koyaanisqa...

[2] https://www.damninteresting.com/the-baader-meinhof-phenomeno...

impressive. surprisingly got something to watch just now

anyone know where I can get the dataset used?

IDK, possibly the worst UX I've ever interacted with.

It's explicitly experimental. I don't think it had the goal to be "good UX" as much as "interesting exploration".

https://themoviepicker.netlify.app

It looks like a fun project. It may have not been designed to comply to the highest UX standards but to try and break things.

Idk why people are always so harsh online. This is clearly stated as "experimental".

Experiments can fail. It's ok for them to fail. And while this one looks pretty cool, it's not usable, at all. Hence the comment.

Reply on news.ycombinator.com