Hello Everyone,
I am Saheed, a university of waterloo grad from Toronto, and also a person passionate about visual impairment, as someone who's struggles with my own disability, I relate to other's.
I created an app that acts as a screen reader for videos. It understands the context, characters, item of clothing and so on. You can use it to watch movies, TV Shows, YouTube, and even twitch VODs.
Here is a link of it in action, for the trailer of The Fault In Our Stars
I also tested it on the most popular YouTuber to see if it can handle his content and I think it does a good job.
The app has a free trial for those who want to give it a shot.
I would certainly love your feedback.
Here is the chrome webstore link: https://chromewebstore.google.com/detail/ibobehoeobnelieehpkjpmccihfopjjj?utm_source=item-share-cb
Comments
So the idea is interesting but...
You have to press describe each time, that takes away from the user experience, how do we decide what is worth describing? Also it's very analitical which has a clear history slant.
I was looking at mississippi god damn, the description went something like this.; 'Nina Simone, sits at a piano, her face displays the frustration of the time...' Yeah that tels me nothing, who's this Nina Simone? There's some black lady singing, is that this Nina Simone? Is she playing the piano, is she the one singing and playing? What i'm getting at is this is too analitical. What does she look like, what is she waring, what color is the piano, what's around the piano, what's the piano seat like, is it short? long?
Do you see? At the moment it honestly says a lot of nothing and no blind person will pay for this.
request for explanation about this screen reader
Hello,
I believe that, because I do not speak English, I did not fully understand how this screen reader works. Even after watching videos of the screen reader in action on YouTube, I still could not understand it, likely because I do not speak English. Could you please explain better how it operates? For example, is it able to provide audio description for videos on YouTube?
I would also like to know how I can activate and properly use this screen reader.
Additionally, I would like to suggest that, if possible, the screen reader be made available in Brazilian Portuguese, as this would greatly help users like me.
Thank you in advance for your attention.
Thank you for the feedback
Thanks for the raw feedback, Brad. You are actually highlighting a known flaw with large language models—they love to infer 'vibes' and historical context instead of just giving raw visual data.
You're absolutely right that users want physical descriptions (clothing, skin tone, environment), not a history lesson on the 1960s. I just updated the AI prompt on the server to strictly enforce 'objective visual descriptions' and banned it from interpreting abstract emotions.
Regarding the manual button: I know auto-description is the dream, but current browser DRM and the speed of AI means live, non-pausing auto-AD is technologically impossible right now without dropping massive amounts of dialogue. FYEO is designed to be an 'on-demand' tool for when you specifically want a deep-dive on a scene. I appreciate you taking the time to test it!
Guilherme
Hey Guilherme,
It is available in Brazilian Portuguese. Just open the FYEO extension pop up and you can select languages. To explain how it operates. It is basically a screen reader for media. It works on Netflix, Youtube, Prime, Crunchyroll, Disney+, Twitch, Internet Archive, and Crave
To Brad
FYEO has an auto-classification engine for YouTube. Because this was a 1965 civil rights video, the AI flagged it as an 'Educational/Historical Documentary' rather than standard entertainment. When it goes into Academic Mode, it shifts from standard Audio Description into analyzing the historical/educational context of the frame.
Thanks for the replies.
At the moment i'd not use this and I still say others wouldn't either honestly. They'd be sitting their pressing the button every couple of seconds and that doesn't seam like something I think peple would use.
Have you gotten other blind peoples feedback off of applevis? Cause at the moment the addon is a bit clunky, for example once you turn on captions, you have to go to the bottom of the screen to find the describe button but a blind person who is new to navigating the web would have no idea and would get frustrated.
If I had to rate your addon, I'd rate it a 2 out of 5, the idea is interesting but we aren't there just yet for AI audio description without prompting the AI each time.
So its paid?
I will definitely check this out, but I was asking about the payment because I would like to know what we are paying for. I mean, as of today, wibe coding a webstore extention like this is a 30 mins job; no coding knowledge/experience needed. 3-4 months back, much before opus 4.7 or gpt 5.5, I remember making for myself a captcha solver extention in under 20 mins. I posted it nowhere because I was like well, anyone could make this now if they wanted; no big deal. I feel this extention is at that level; and if the payment is for the LLM use itself, the better way is to offer a choice to the user to use their own model and API key if they want to.
good try
I think it is a good one.
if it can work out of chrome then can describe the video in gaming ...
it will be so cool!
Good initiative, but I would pay if the extension provided audio
I liked the screen reader; I found the idea very interesting and useful. However, I think having to click the “describe” button all the time during videos is quite inconvenient and not very practical.
I would like to know if you plan to add a feature where the screen reader provides audio descriptions of videos in real time, without needing to keep clicking the “describe” button.
If this feature is implemented, I would definitely be willing to subscribe to the service.
To Guilherme
I do plan to add that feature -- when the AI is ready to do it. For now, this is the best the AI can do without being incoherent. Thank you for the feedback.