A window to every image on Discord.
PixelRoom describes images, GIFs and short videos for blind and visually impaired people. Type one command and hear what everyone else is seeing. Only you get the answer, in your own language.
- Mira Look what I found at sunset!
- You /describe
- PixelRoom APP Posted by Mira, just now. A pink and purple sunset sky over a green hill, with a bright yellow sun low on the right. The mood is calm and warm. Only you can see this.
How it works
PixelRoom stays quiet in the background. It never posts on its own, so it doesn't fill your channels with messages.
-
Someone posts an image
A photo, a screenshot, a meme, a poster, a GIF or a short video clip.
-
You type /describe
No need to find the message. PixelRoom picks up the latest image in the channel.
-
Only you get the answer
A short description, written for screen readers, in the language your Discord is set to.
Commands
Every answer is private. Other people in the channel don't see that you asked.
- /describe
- Describes the latest image, GIF or video in the channel.
- /describe detail: long
- A longer, more detailed description, for when the short one isn't enough.
- /describe count: 3
- Describes the last few images in the order they were posted, up to 5. Useful when you come back to a busy conversation.
- /describe back: 2
- Describes an earlier image. 2 is the one before the latest, 3 the one before that.
- /ask question: …
- Ask anything about the latest image, like "What colour is her dress?" PixelRoom repeats your question before the answer.
- /readtext
- Reads out all the text in an image, exactly as written. Good for posters, rules and announcements.
- Describe image
- For a specific message: open its menu, then choose . With a keyboard, focus the message and press Shift and F10 to open the menu.
Made for people who use screen readers
Every choice in PixelRoom starts with one question: what would make this easier to hear?
-
Short and to the point
The main idea comes first. Game screenshots describe the scene, not every button and health bar.
-
Your language, automatically
Answers come in the language your Discord is set to. No setup needed.
-
Text read exactly
Text in an image is written out word for word and never translated, so you get the original.
-
GIFs that move
PixelRoom watches the whole GIF or clip and tells you what happens in it, not just one frame.
-
No guessing who people are
Real people are described, never named from their face. Characters from games and films can be named.
-
Plain text only
No emojis, symbols or formatting that a screen reader would read out loud.
What happens to images
PixelRoom only looks at an image when someone asks it to. Images nobody asks about are never sent anywhere.
When you ask, the image is sent to Google's Gemini AI, which writes the description. PixelRoom doesn't save the image or the description.
Read the full details in the privacy notice.
Questions
Does it work with my screen reader?
Yes. PixelRoom answers in plain text, which every screen reader can read: NVDA, JAWS, Narrator, VoiceOver and TalkBack.
Can other people see that I asked?
No. The answer appears in the channel, but only you can see it, and it disappears when you reload Discord.
What if the image already has alt text?
You can still ask PixelRoom. It describes the image anyway, so you can compare it with what the person wrote.
Can the description be wrong?
Sometimes. The AI is good, but it can miss details or misread something. If you're unsure, use /ask to check a specific detail.
What does it need to work on my server?
Permission to view channels, send messages and read message history. It doesn't need to manage anything or read anyone's private messages.
Open the window on your server.
It takes a minute to add, and it stays out of the way until someone needs it.
Add PixelRoom to your server