FullStackToolkit VS Parlune

Parlune's answer:

Two things, and they work together.

First, Parlune creates its own transcript from the video's audio instead of relying on whatever subtitles happen to be available on YouTube.

That matters more than it sounds. We looked at 392 YouTube videos across ten languages and found that only 5.1% of Spanish videos had human-written subtitles, while 30.8% of Arabic videos had no text in the language at all. And even when auto-generated captions exist, they're often poorly punctuated and difficult to study from.

Parlune turns the actual speech into clean, properly punctuated sentences, whether the video has subtitles or not.

Second, Parlune doesn't stop at the transcript. Every word and sentence is aligned to the audio, so anything you save can be reviewed as audio. Your saved vocabulary feeds into 11 short review games covering listening, recall, recognition, comprehension and spelling — all using the original speaker's voice.

So you're not just memorising a translation or recognising a word on a flashcard. You're repeatedly hearing that word in real speech, spoken by a real person. the original speaker's voice rather than a generic recording — so review is another encounter with the same real speech, not a text flashcard.

Parlune's answer:

Because Parlune delivers high-quality language learning materials with a focus on audio-first review.

Tools such as Language Reactor depend on the subtitle track available in YouTube's player. Trancy uses existing subtitles on its free plan and offers AI transcription on Premium. Parlune takes a different approach: every video is transcribed from its audio as standard.

That means a video with perfect professional subtitles and a video with no subtitles at all are treated the same way. You choose something you actually want to watch, and Parlune turns it into study material.

The other big difference is what happens after you save something. Your vocabulary comes back through spaced repetition as short audio-based games using the original speaker's voice. The aim isn't just to recognise the word again — it's to get better at understanding real speech.

Parlune's answer:

Parlune is primarily for intermediate language learners who have outgrown beginner apps and want to start learning from real content.

Usually, that's someone who has worked through Duolingo, a textbook or a structured course, can handle relatively simple material, and has reached the point where textbook conversations no longer sound much like the language they actually hear.

They're often already interested in immersion and comprehensible input — things like Refold, Dreaming Spanish or sentence mining — and don't necessarily want another course or curriculum. They want to take videos they're genuinely interested in and turn them into something they can study.

Because Parlune currently translates into English, learners also need to be comfortable reading English.

Parlune's answer:

Parlune started with a much simpler idea: take YouTube's subtitles and turn them into language-learning material.

Then I started trying it with real videos.

The problem is that a large proportion of YouTube videos either have no subtitles at all or only have YouTube's automatic captions, which are infamous for being inaccurate, and completely unpunctuated - making them unsuitable study material when learning a new language.

Rather than assume this was a problem, I measured it. We sampled 392 YouTube videos across ten languages and found that only 5.1% of Spanish videos had human-written subtitles.

That changed the direction of the product. Instead of depending on subtitles, Parlune would transcribe the video audio directly using the leading speech to text LLM's to ensure high-quality transcriptions.

It's a more expensive way to build the product, which is one reason Parlune is priced by hours of video rather than offered as a free tool. But it means the quality of the study material doesn't depend on what YouTube happens to provide for a particular video.

Parlune's answer:

Parlune is built around a fairly straightforward stack:

• React single-page application with Webpack and Tailwind CSS
• Node.js and Express API
• MongoDB
• Astro for the marketing site
• AI speech-to-text models for transcription
• LLMs for sentence segmentation, translation and word-level enrichment
• yt-dlp and FFmpeg for audio acquisition and processing
• Vercel and Cloudflare for hosting and infrastructure