When I uploaded my subtitle file, the generated course’s text is different from my original subtitle file’s content. It adds some unnecessary words or skips a few words. I’m suspicious that it is because of the internal ‘AI-helped’ feature. I want to turn the feature off but I can not find such an option in setting or import page.
Why is this unnecessary feature mandatory? Please allow users to turn it off manually, it is really annoying.
Can you please send the file you imported to support@lingq.com? We will investigate it.
If LingQ uses some LLM with a prompt to make a non-formatted transcript neat version that is easier to read, consider trying this prompt that I am using personally, It works better than LingQ’s original importer.
You are an expert in restoring punctuation. Your task is to transform the following text into complete sentences while preserving every original word and phrase.
Please follow these steps to complete the task:
1. Convert the input text into sentences according to these guidelines:
a. Preserve every original word: Do not add, remove, or change any words from the original text.
b. Create proper sentences with appropriate punctuation.
c. Form logical paragraphs by grouping related sentences. Start a new paragraph when there's a shift in topic or speaker.
d. Apply formatting: Use no line breaks within paragraphs and separate paragraphs with a blank line.
2. After completing the conversion, review your work to ensure:
- Every original word is preserved.
Remember, the key is to improve readability and structure while faithfully preserving every original word and phrase. Pay extra attention to ensure no words or phrases are skipped in the process. What you only can add to the original text are punctuation and new lines.
A lot of scripts imported from YouTube have different starts and ends than the original subtitles. For example, some sentences are separated for a natural connection, or sentences are ended with ‘…’. Sometimes, parts of the subtitles (several minutes) are missing entirely (this usually happens at the beginning).
This is a common problem that I’ve experienced when importing dozens of videos, so you might be able to reproduce the problem on some random video you want.
This is one of the videos that suffers from the problem.
and this is the fragment that shows the problem.
in an extremely high-profile deal Samsung was supposed to make these special memory chips for NVIDIA chips crucial for nvidia's AI efforts but in one of the biggest blunders of the Year reports indicate that Samsung completely fell flat their chips couldn't reach the strict requirements for heat and power efficiency that Nvidia required and it was a tough blow falling short of a major client like this comes with consequences and for Samsung this was costly meanwhile competitors like SK HX have sweeped in to take advantage and now hold a strong lead in the high bandwidth memory market and now they're currently the exclusive supplier of these chips to Nvidia if you take a look at this graph for the high bandwidth memory Market just at a glance we can see that Samsung's flat white line is consistently at the very bottom trailing well behind its smaller competitors in the high bandwidth memory space Samsung knows they're falling short so they published a recent apology acknowledging M targets and struggles in high performance AI memory investors were further disappointed with Samsung's latest quarterly profit it was 900 million below expectations they still did make 6.78 billion and that's nothing to sneeze at but Falling short by nearly a billion doesn't exactly make investors jump for joy Samsung was in panic mode and vice chairman juny hians recent statement blatantly admitted that their failure to meet Market expectations has fueled concerns over Samsung's technological competitiveness and future Direction this quote unquote crisis as Samsung management calls it even led the company to enforce a Six-Day work week for its Executives in an attempt to regain lost ground in short Samsung who manufactures chips were caught napping as AI chip Hardware took off and investors aren't happy
When LingQ importer works badly I regenerate the lesson manually. I’m sharing how to do this for the users experiencing the same problem with me:
- I extract the transcript from YouTube with this website https://www.youtube-transcript.io/
- Concat broken sentences and make the script neatly to make them look like complete text using LLM (GPT, Claude, Gemini) with the prompt I attached.
- Edit lesson > Regenerate Lesson > Paste the complete transcript > View Lesson.
- Although the new transcript does not provide synced highlighting with Youtube video, It works nicely.