Hi
Reaction
Comment
Share
New Feature Request 💝
Publish now
Visual context awareness function
When summarizing YouTube or PDF files, it appears that the release only uses the voice recognition content or subtitle files from Whisper for videos, and only text data for PDFs.
In the case of YouTube videos, screenshots are taken of important parts and added to summary notes, but it seems that the context within the screenshots or videos is not recognized.
In the case of PDF, it only summarizes the text, but if there are images or diagrams in the PDF, it seems that it does not recognize the context.
Claude used to only extract and summarize text when you uploaded a PDF, but after a recent update, a new function was added that actively utilizes vision capabilities to recognize context.
I know that the current release uses GPT-4o, and if you actively utilize the vision features in the model, you can expect better summary notes. (It might cost more, but... I think it would be great if it existed... ㅠ)
1 Comment
Reaction
Comment
Share
It would be nice if there was a feature where the current playing video time would be synchronized with the summary timeline and highlighted. I keep looking back and forth to see which part of the summary the video is in.
Basically, I would like the script to be exposed below the video. It might be a bit difficult on mobile, but on large screens like pads or computers, there seems to be space for the script subtitles to appear below. It would be icing on the cake if the currently playing part was synchronized and highlighted.
1 Comment
Reaction
Comment
Share
Ability to upload files directly from the cloud
When uploading lecture videos or PDF files, I understand that only files that exist locally on the system can be uploaded.
It would be nice if it were possible to add files directly from the web by linking to web hard drives such as OneDrive, Google Drive, and Dropbox.
And, if it is technically possible to summarize the footage by loading it directly from the drive rather than copying it to the release,
In other words, if you register for a release in web format with Google Drive, it seems possible to prevent it from being deducted from the release's own file storage capacity.
(Probably only the Whisper capacity will be deducted)
1 Comment
Reaction
Comment
Share
When summarizing a PDF, if there is a table of contents in the file, it will be summarized according to the table of contents.
I'm really enjoying the PDF summary feature.
However, I think there is a problem that the table of contents in the summary note and the documents within the PDF document are different, so different contents are mixed in one table of contents.
Is there any way to automatically recognize the table of contents within a PDF, or to allow the user to input a specific table of contents directly?
2 Comment
Reaction
Comment
Share
When translating and summarizing, I would like to be able to allow the user to specify specific words so that the translation-summary can be done.
For example, in a CSV file, column A contains the words to be translated, and column B contains the translated words specified by the user.
For example, in a CSV file, column A contains the words to be translated, and column B contains the translated words specified by the user.
Also, I would like to be able to specify multiple functions like word translation list 1, word translation list 2, etc., so that users can specify word lists by the area of the papers they want to see (statistics, economics, management, sociology, cultural anthropology, etc.) and have translation-summary-AI conversation.
1 Comment
Reaction
Comment
Share
The viewer UI for audio files is too inconvenient
The video summary function is the main feature, but the voice recognition and script creation function is just as good as other services, so I upload lecture recordings and use them.
But in this situation, I wonder if the viewer is designed around the video, so the audio file leaves space for the video even though there is no video. How can this not be improved...?
+ When you upload a video file separately, the player UI is fixed by default, which is not bad to use, but it is still inconvenient in many ways...
1 Comment
Reaction
Comment
Share
Speaker identification function
When I summarize a podcast video or upload a transcript of a recorded discussion, it seems like I can't tell who's speaking.
I was thinking of releasing it as an all-in-one summary platform and moving voice recordings to the release, but the inability to distinguish speakers is a bit fatal...
I think I would use it more often if it had speaker identification, like Daglona Clover Note.
1 Reaction
1 Comment
Reaction
Comment
Share
Ability to summarize video using query string
There is no major inconvenience when using it on a desktop. When watching a video on a mobile device and want to summarize it,
Share - Copy Link - Move Release AI Link - Paste
I think it would be useful if you added a feature that seems to shorten this process.
If you create a shortcut on your iPhone
Share - Click on the shortcut you created
It seems like the process would be this simple.
3 Reactions
1 Comment
Reaction
Comment
Share
Added trash can feature
When deleting a note, a pop-up window appears and you go through a process of confirming whether or not to delete it. However, when deleting multiple notes, you often press the confirm button without thinking because you are used to this pop-up window.
The Delete Collection button is also close to the Delete General Note button, so I press Confirm without thinking...
Yes, I blew away 14 summary notes like this...
I wish the Recycle Bin feature would allow for some level of restoration...
It's so unfair....... ㅠㅠㅠㅠㅠㅠ
1 Comment
Reaction
Comment
Share
AI chat window obscures the source
When you turn on AI chat, the window will pop up in a floating manner at the bottom right of the screen.
However, when using sources in PDF or text format, the chat window covers most of the source because the source window is on the right.
For sources such as PDFs or texts , a view that simultaneously shows the AI chat and the source is preferred because you need to check the source to see if the AI is making any strange noises .
However, in the current UI, it is difficult to see these two together.
Here are some solutions I can think of:
Make the chat window move around the screen
For now, the chat window remains fixed at the bottom right of the screen.
If we could allow users to move this window around the screen, it would give them more room to improve the situation themselves, such as putting the chat window on the left and the source on the right.
Make the chat window resizable
Actually, it doesn't seem like a bad choice of location to keep it fixed at the bottom right. The only problem is that the chat window takes up too much space.
If users could adjust the chat window size on the fly, that would seem like a good compromise.
Provide an "Open in new window" option
An option to open in a new window would also be nice. Since users with dual monitors or split-view can use the multitasking features provided by the OS itself, I don't think there would be much of a difference in sources if this feature were available.
If hiding the source is the problem, another solution might be to provide an option to open the source in a new window .
See more...
1 Comment
Reaction
Comment
Share
Summary Queue Function
I understand that the current release does not provide concurrent summarization functionality.
I think it would be nice to have a convenience patch that would allow me to queue summaries for the materials I need in advance when I don't necessarily need the quick summary boost feature.
Of course, this feature will become unnecessary once the simultaneous summary feature is introduced, but I think this is more realistic and economical.
2 Comment
Reaction
Comment
Share
After summarizing the document, I found that there is a function to click on the expected question in AI chat in the right pane.
It seems that expected questions are generated better than expected and answers based on those questions are also very good.
Rather than the context written in a blog post when summarizing a document,
Questions recommended by Chat AI
It creates context
More compelling questions and answers are completed
I think that if you only select the questions you need, the answers will be organized into good results.
However, you can only copy this one question and one answer at a time.
I can't download the entire Q&A in document format or copy the entire document (at least this please)
See more...
1 Comment
Reaction
Comment
Share
I wish it was easy to adjust the location of collections (categories). I want to move from the basic collection to another collection, but it would be nice if there was a way to move using the right cursor. I'm organizing my thesis...ㅠㅠ I really need this feature. I'm dividing documents by category.
1 Reaction
1 Comment
Reaction
Comment
Share
When using summary content, there are many cases where proper nouns such as names are recognized incorrectly and must be corrected one by one.
It would be difficult to improve the accuracy of AI, but it would be nice if basic functions such as text replacement in document editors were easier than AI.
Since it is an AI service, it would be even better if there was a function to recognize incorrectly entered names and correct them to the correct names by giving the AI a user dictionary.
It would be nice if there was an option to highlight the text to be modified like vscode and have the editor just look at it and press enter to modify it.
It would be difficult to improve the accuracy of AI, but it would be nice if basic functions such as text replacement in document editors were easier than AI.
Since it is an AI service, it would be even better if there was a function to recognize incorrectly entered names and correct them to the correct names by giving the AI a user dictionary.
It would be nice if there was an option to highlight the text to be modified like vscode and have the editor just look at it and press enter to modify it.
2 Reactions
2 Comment
Reaction
Comment
Share
Share
D
Daewhee_Kim
LLM utilization other than GPT-4o
I would also like to have the ability to select models like the Claude 3.5 Sonnet, LLaMa 3.1.