TL;DR: I built a web app called LangGlitch that aggregates comprehensibleāinput videos (and soon podcasts/graded readers) for multiple languages, starting with German, Vietnamese, and English. I want it to become a oneāstop place for comprehensible input for every language, including small/obscure ones, and would love your feedback and language requests.
Hey folks, Iām Stefan, a guy from Germany who loves travelling and getting lost in new cultures and languages. In school I was terrible at languages and grammar never really made sense to me, but at some point I realised I had somehow become fluent in English just by playing games and watching YouTube in English.
Looking back, what worked for me was basically ācomprehensible inputā: content that I mostly understood and actually cared about, consumed for fun rather than as āstudyā. These days I always try to learn new languages that way, but I kept running into the same problem: unless youāre learning something huge like Spanish or Japanese, good comprehensibleāinput content is scattered and hard to find.
So I decided to build something for myself and ended up turning it into a proper project: LangGlitch ā a little web app that aggregates comprehensibleāinput videos for language learners. Right now it supports German, Vietnamese, and English, with playlists grouped by difficulty, tags, and creators. You can sort for āeasiestā, filter for topics you like, and then just watch your way through material instead of hunting for the next decent video.
Iād love for LangGlitch to eventually cover every language out there, including the really small and āobscureā ones, so it can be a genuine oneāstop place for comprehensible input. If thereās a language youāre passionate about and want to see added sooner rather than later, tell me in the comments or message me and Iāll do my best to prioritise it.
Iāve just put it into free open beta, so anyone can sign up and play around with it. Iām planning to add more languages over time, plus podcasts and graded readers, and if it ever makes enough money to pay its own bills Iād love to commission new comprehensibleāinput content for underrepresented languages as well.
If you try it, Iād really appreciate honest feedback: confusing UI, missing features, annoying bugs, anything. You can leave comments here, DM me, or join the Discord (linked on the site) and yell at me there. Screenshots in the comments so you can get a feel for how it looks.
Wow, thanks for the amazing feedback! This helps a lot!
Your concerns about the Landingpage are valid & yes its pretty much a layout copy of cjjapanese. I myself am not very good at marketing & design so thatās why I just started with something that already works. The Landingpage will get a full redesign after some feedback (I will for sure highlight the currently available languages!).
On your thoughts of the infinite scroll, I understand your sentiment but I think this is also highly subjective. I could probably add a setting or something similar that changes form infinite scroll to paging but that would be a pretty low priority at the moment. Iād rather improve the infinite scroll in away that it isnāt annoying, so would it be better to not load randomly & order them by Difficulty as the default option?
A Watch-Later option is something I will do for sure but its also not the highest priority but a nice feature that will for sure come.
The Tag loading problem might happen because I didnāt scale up the server very high at the moment for cost saving reasons during the test phase. This should resolve itself when I scale up the Server & databases at a later point. I agree that there are a lot of tags but there are also a lot of different Videos on the platform already & they will only expand. Maybe I can consolidate some at a later point & remove the ones that only have very few videos.
I like the Calendar progress tracking suggestion. I will add that!
Hmm, I see your point that you could just use my platform to discover the Videos & then watch them on Youtube but that is a general problem I guess. But what the platform is offering isnāt necessarily the content but aggregation, discoverability & curation and some people just might not value that. So if I offer unlimited watch-time in the free tier it might not be sustainable as a business because Progress tracking alone isnāt a selling point in my opinion. The problem on the Free tier also would be that it would be Ad-supported & that doesnāt really make a lot of money unless you go overboard with them. Currently you make $0.00025 to $0.002 per banner ad impression which isnāt a lot & I only planned to add some ads into the Video browse page & a small banner into the Videoplayer so I donāt think I would be able to recoup even only the cost on a free tier user. But nothing is decided yet so I might just have to go with it anyway to get people to use the app.
Yea subscription fatigue is a huge problem right now, especially for SaaS businesses. Thatās why i would probably aim between 2-4$/month for the premium version. Also Dreaming & CiJapanese are for sure competitors but only in one language so I think there is still a lot of value in my Service. Also I donāt see the free Youtube Videos as a big competitor because the core of the business isnāt the content itself (not yet at least, I hope I can produce some later).
Right, it is for sure just my opinion and probably not a very popular one.
Loading randomly is still an important thing, I think. Iād have liked to not be surprised by it, but I do think itās important to have a way to get a random selection of videos within a difficulty tier.
Is there a way you can load that in the background as the rest of the page loads, and/or cache it? Point taken though, not something to sweat too much if itāll self-resolve.
A lot is not necessarily bad, just keep in mind the keyword is curation. Constantly evaluate to see if the tags are helping users find what they want, and make sure the tags on the videos are highly accurate and relevant. That probably does lead to more.
I think Youtube really, really is your biggest competitor though (Youtube itself as a front-end, not the videos on it). It may be a general problem, but itās the biggest one you have to solve - why watch videos on this site instead of going to Youtube? I think the embeds always have that big āwatch on Youtubeā button for them to conveniently leave, after which theyāre in Youtubeās discovery pipeline with no restrictions. Even if they can no longer see recommendations on LangGlitch, they learned the name of at least one or two channels so they know where to look. YTās recommendation algorithm takes over from there, plus most good CI channels already have nice playlists organized by level.
To be honest, I just had this kind of experience with Dreaming Spanish the other day. Realized they had a website, started watching videos there, then after a day or two got told ānoā on something and immediately left. Sorry, I donāt know how to turn this around and make it very constructive, I guess Iām just giving my thoughts as a learner.
I guess the way Iād look at it is you have to offer a lot of value up front, then gradually alter the deal over time. That sucks for customers, but I guess itās a way to make money. Thatās the playbook for most online services - offer a sweet product at free or ridiculously low cost, get people to love and depend on it, then degradify and turn the screws on price. As a human not backed by a VC or megacorp though, I dunno that thatās viable. (Sorry againā¦ this paragraph sounds way more cynical than I wanted it toā¦ Anyway, strong note to not take any business advice from me, I donāt know anything about monetizing a product or running a company.)
Yep, going multilang is a great differentiator here. Small note though that Dreaming does Spanish and has started doing French, and sounds like they plan to do more in the future, so sharks in the water there. But I think youāve got an angle, targeting less popular languages and curating from lots of channels.
Again, best of luck. I think itāll be very challenging to make it work, but it probably can. Tagging and organizing tons of videos and making them easy to find is really admirable. Two other random possibilities that come to mind that could be cool: crowdsourcing tags or most relevant tag rankings, detailed transcription search and stats per video
Curious also, whatās your methodology? Do you run everything through an LLM pipeline, grab tags directly from tags on YT, manually review?
Your Feedback about Youtube is pretty valid tbh. So yea no Idea on how to handle that problem but I just hope people like the platform enough to use it over Youtube. It worked for DreamingSpanish in the beginning (before they had premium videos) as well so it could also work for me. I guess we will see.
And yea the plan to enshitify overtime is not really something I wanna do. I just wanna over value at a reasonable price so people just donāt mind paying for the convenience. Also I donāt think is even a valid way if you donāt have VC money to burn.
And yes I saw DreamingSpanish changing to Dreaming.com & them offering French now but I think their business model is just different. They focus on extremely high Quality Content that they have control over & that makes them expand slower in to new languages. So for the meantime it should be fine.
About your two Ideas:
My Methodology rn is the following: I use a tool I created to scrape/download all the Video Information from a Channel that I want to Integrate into the platform. This data gets put into a google docs sheet & then let an LLM Model (n8n) run over all the data to change Title & Description. After that I manually Double check titles & descriptions. Now I assign tags & Difficulty and other internal values. Last step is importing the Google Sheet into my Platform.
Iām currently working on a fully integrated process inside my Admin tool to have less manual steps (copying back and forth, manual Import, etc.)
Always glad to hear that! Hope I didnāt come across as accusational, just kinda where my mind went trying to put on a business hat. Scary thing (maybe not for your service, but for many) is that fair-deal services have to compete with those that offer unsustainably high-value services cheap/free to get people in the door (I mean, I guess Youtube is a good example of this).
Another musing, might be nice to have videos from other sources too. PeerTube, Odysee, Vimeo, BilliBilli exist. Tiktok has some good stuff, for sure. I donāt think this is worth the trouble though, considering those first few have almost no content.
True, it would be a lot of work. Maybe, depending on your user-base, a tag only gets nominated if some # of users suggest it (say, 2+ for now). Then it goes to the suggestions queue for you to look at it. Maybe new tags, but maybe not even that, just other tags you already have that arenāt applied to any videos. I realize you already have a āsend feedbackā thing though, so thatās probably enough right now. Also, this could involve your Discord - maybe a channel automatically gets posts sent like āShould this video have the xyz tag?ā and people react thumbs up, thumbs down.
Rankings - my thinking was a few tags could maybe show under the video when browsing. Or searching by tag could return ones where the tag is most relevant.
Glad to hear the tags are human added. LLM only approach would scale better, but produce silly nonsense sometimes and make the site less fun. I do think if you have any LLM generated content (titles and description) thatās user facing, that should probably be disclosed somewhere in an About or FAQ page. For the difficulty though, that can probably use some input from automation (not even necessarily LLM, just good old fashioned code) - might not be too hard to figure out words per minute, % of common words, etc from the transcript and run time. But difficulty values from the channel that created it are probably more accurate, and I imagine most of them say what theyāre targeting. And analysis wouldnāt capture things like visual aids.
Iām was thinking about this. I have to see which of these platforms allow embedding. I know TikTok does & BilliBilli doesnāt but no idea about the rest.
I like this in general. I decided against the first point because it kinda overloads the UI a bit. Maybe if I find someone who can help me with design I can revisit this. The last part sounds like something I should do. Like ordering the Videos that have the tag by relevance sounds like a decent Idea.
Well yea a pure LLM approach would be for sure faster & scale easier but you just canāt trust AI output enough so you need supervision. I actually tried to get a general sense of the difficulty based on LLM output but that took me nowhere so I had to revert back to do it manually. Most of the time I use the Difficulty the Creator has set but a lot of the times there is no difficulty anywhere & thats where I have to evaluate it myself. This is actually the hardest part because I donāt speak a lot of the languages so Iām depended on friends helping me out there.
This is actually harder than you would think. I had the same Idea after my LLM approach & it turns out this is hard as shit. Even linguists that studied a language struggle with this. Also getting data setās for the most common words might be easy for English, Spanish, etc. but not for the more obscure/rare ones. In my research I found a service that offers this for English but itās still mostly for written words TextInspector. Having a transcript of spoken words makes it even harder because there is no punctuation & spoken word is not always correct or uses slang. Like this whole topic is extremly complex & creating the Datasets could be a whole business by itself.
Also I added a Watch Later feature. Feel free to check it out & tell me what you think about it.
That 100% makes sense. Anything involving language seems to have potential to be a rabbit hole problem.
That was quick! Looks good. I was a little surprised to see the āvideos like thisā sidebar when going back to watch them, instead of the playlist sidebar (since these are accessed from playlist page). But I donāt necessarily dislike that either.
Couple other things that I noticed while trying that: the āSuperbeginnerā label feels hard to read, might need more contrast. When watching a video, the difficulty tag doesnāt seem to be anywhere. When browsing the video list, hovering over difficult expands and shows a number - but clicking still just goes to the video. When watching a video, tags arenāt clickable - Iād like to be able to click those to go to a search filtered by that tag. Creator is also not visible on the video player page, would be another good thing to be able to click.