opal-general・✨

[bleak anchor]

**## Introducing: Opal **Opal is a new experimental tool from Google Labs that lets you build and share powerful AI mini apps that chain together prompts, models, and tools—all using simple natural language and visual editing.

Try Opal for free at opal.withgoogle.com and be sure to share your thoughts with the Opal crew right here in this new channel! Whether you've found a bug or want to share your designs, this is the place to be

[ancient vector]

I'm in love with it! I'm new media artist Orkhan. Would love to get early access

[wise oasis]

Interesting, I'll note the landing page moves directly from a landing page that seems to decribe "How to Build an App" to a page with my apps and templates, which makes it hard to learn how to build an app

Going to the blog seems to get one access to this information:

https://developers.googleblog.com/en/introducing-opal/

[quaint eagle]

that's really good feedback! would you like to see the video when you're starting a new app?

[wise oasis]

I think I'd like a link to the video, and any other documentation. I don't mimnde if the opal website takes me to my project, it's just a little weird that it shows me the landing page very briefly, then takes me to my projects. I may be experiencing this because I dived into mucking around with a remix, and only later wanted to go back to the landing page watch the video.

[thorny grotto]

I found a bug

the theme editor doesn't work!

nothing generates

oh wait it did generate

but it takes a lot of time to appear

[quaint eagle]

yeah, it's a bit slow

[thorny grotto]

after the genewration is done

[thorny grotto]

oh

also

what are the limits on veo there?

[dense totem]

Thanks for the feedback and the follow up @wise oasis. As Dimitri said, this is a great point. We are actually addressing this as we speak and should be able to release an update with the video now available in the signed-in, page (where you see projects, etc) and when you click on Create New

[quaint eagle]

@wanton blade makes me wonder if we should make it somehow work in parallel or something

[wise oasis]

Wow, I was able to get it to create a generic roleplaying app in a minute or two with is simple prompt. Can't wait to dig in a try creating a custom setting by uploading assets.

https://opal.withgoogle.com/?flow=drive:/1TQYhvVdhk9piz5GS4XeQQmhOMjA4AsPI&mode=app&shared=true

[dense totem]

There is a limit on the number of videos you can generate, which could be adjusted depending on traffic and global usage. Roughly speaking, you should be able to generate ~2-3 videos in a minute.

[thorny grotto]

oh damnn

thats a lot

because we can only get like 3 videos/day with the pro plan

[dense totem]

Thank you @wise oasis. Glad to her you were able to create th is so quickly! I just took a look and it's actually something I could use! 🙂 Can't wait to see what other apps you make and keep us posted about your experience with the product

[thorny grotto]

Opal doesnt appear in labs.google site lol

instead I see a new experiment 👀

something called archives

[quaint eagle]

we went for a low-key launch 😄

[thorny grotto]

ohh

is this a bug

haha

[quaint eagle]

unexpected errors are usually bugs 😆

if you try again, does it do the same thing?

[thorny grotto]

ohh lol xD

[thorny grotto]

lemme try

[dense totem]

Can you share the full text of the app description you used here?

[thorny grotto]

sure

[thorny grotto]

I want to build an app where on entering the name of a chemical compound, we get a mini description of it, a preview of the 3d structure of it (built with coding in HTML) and a short video showing how the Chemical Compound is used in Real Life

[thorny grotto]

yess it does

[thorny grotto]

(dont know if it can make this as it's totally new for me 😆 )

[dense totem]

we're trying it now as we speak

[thorny grotto]

thanks!

[dense totem]

we can reproduce the bug. Debugging it now...

[thorny grotto]

thats great 😁

[dense totem]

Can you try this app description instead?
"build an app where on entering the name of a chemical compound, output a mini description of it, and a preview of the 3d structure of it (built with coding in HTML) and a short video showing how the Chemical Compound is used in Real Life, all rendered into the HTML"

[dense totem]

The reason your first app description didn't work is that it read like it needs to create multiple outputs, vs in my version, I put 'all rendered in one HTML'

[thorny grotto]

Oh I see

It did take the input

did you get the output when you tried the prompt?

or i dont know if its the processes

[elder wigeon]

the generating video takes a longer time.

since veo3 is....slow

[thorny grotto]

oww

I understand

so it's the processing thing or the loading screen

[wanton blade]

Can you tell me which browser you're using?

[dense totem]

Right, it is a bit slow.

[quaint eagle]

check this out. Now, when you start a new app, it will give you a link to the video

[thorny grotto]

Im using Zen (a firefox based browser)

[analog karma]

Yooo

[thorny grotto]

https://opal.withgoogle.com/?flow=drive:/1wos3RKVOIZ1k9KHQduffszavJUE6d8H7&mode=app&results=15B1VZCLPmZKtthAsPKlsup0Pn1CNOsDS

my first opal! 💖

but it had some issues with the display lemme tryna fix it

[elder wigeon]

https://opal.withgoogle.com/?flow=drive:/1JfPaNP__Cw8TvMVVLJqutyzuLNCXtuqa&mode=app&results=161y6V7AMUjz0qe3q-tvtkMg30qnFCzP1

this is the output when i tried your prompt 🙂

[thorny grotto]

your link gave me this error

[elder wigeon]

oppps

let me try again

[wanton blade]

you will need to publish it first

[elder wigeon]

https://opal.withgoogle.com/?flow=drive:/1JfPaNP__Cw8TvMVVLJqutyzuLNCXtuqa&mode=app&results=161y6V7AMUjz0qe3q-tvtkMg30qnFCzP1

try again?

[thorny grotto]

I got something similar. the output window shrunked to the top left corner and the code preview nor working properly

[thorny grotto]

ohh

I got something like this

[quaint eagle]

this is likely a cross-browser rendering problem. We'll sort it out. For now, use Chrome 😄

[elder wigeon]

hmmmm yeah looks like a browser rendering issue for now.

[thorny grotto]

oww I see haha

lemme see how it looks like in chrome

[elder wigeon]

or chromium 🙂

[thorny grotto]

yea i think so. Mariner also had a similar problem with zen/ff but works with other browsers like arc/edge/comet,etc

[wanton blade]

Okay, just taking a look to see if I can figure out what might work here

[thorny grotto]

thanks!

[thorny grotto]

I tried it on chrome and saw a lot more elements in the site than zen

It worked good except one thing, the struture coded in HTML lol. It just shows me an image

I think it's because Opal doesn't have a built in Code Preview as it's not something very relevant to the tool but others were fire 🔥

[wanton blade]

Okay, I think I found some places where we can improve it for FF-based browsers

[wise oasis]

So I've trying to improve the "text adventure game" to move from a generic game as I made initially, to game that is grounding in a setting I've published. (Designing and publishing TTRPG is a hobby of mine) I've tried linking docs from my drive, uploading a PDF as a asset, or mentioning the link in the "describe what you want to build" box, but doing this seems to fail, or ends up with the app asking the user to link to a game setting document.

Do you have any suggestions on the best ways to share content to ground and inform the app?

[quaint eagle]

yup, the way I would do it is by editing the app you already have and adding assets to it

let me post a screenshot

[wise oasis]

More generally, I'll note that I when I want to create a specfic study guide, based on particular YouTube video, rather than use the generic demo app you have provided, Opal keeps wanting to have the user upload a link to a video. It feels (though this is probably my inexperience) that it really wants to default to asking the user to provide content, while, to my mind at least, what it powerful is if I can provide the content, and have the app build interactions that are grounded in this.

[quaint eagle]

let me know how I can help

[wise oasis]

Here's a screenshot of a project, you can see that I added an asset (in this case a link to a YouTube video, but when I prompt Opal to use that asset to create a quiz, it defaults to starting with that Yellow box that asks the user for the video. I'm just trying to figure out how to prompt or work so that Opal focuses on the asset I have uploaded, and avoids asking the user for what the content the quiz should be about (or the game should be about) or whatever.

I'm sure there is a way to do this, it's just confusing to me what I should be doing

[quaint eagle]

you're almost there.

• click on the yellow box and hit "Delete" (if you're on a mac or "Backspace" if you're on Windows)
• grab the little round handle on the right top side of the YT video and drag it into the the "Extract Video Description..." step

basically, you need to wire the video into the blue box

[wise oasis]

Super, thanks, I'll try that

Yes, that seems to work, now to refine and improve!

[ancient vector]

gm Dimitri, how can I get early access?

[long jetty]

really really interesting....excited to play around with this one.

[quaint eagle]

if you're in the US, you should be able to access already

[quaint eagle]

here's a video of me wiring an asset into a generate step. Hope it helps: https://youtu.be/1LpV_5Qa6KI

[tranquil goblet]

getting this error for all the default gallery apps I've tried so far

[fast oxide]

Is there any way to format the text response from an Opal app? Mine is coming out in large type on a green background, centered and overflowing the text area. Not ideal. I tried updating the query in the rendering step but it made no difference.

[quaint eagle]

yes, updating the prompt in the output step is the right approach. Sometimes it takes a bit of convincing and clarifying

[fast oxide]

Any tips on convincing and clarifying?

Here's what I tried:

Specifically, the webpage should have a light grey background, Arial 12pt black body text, Courier 11pt black code examples on a light blue background, left alignment for all content, and optional clickable links to copy code examples. Headings should be styled as follows: H1 in Arial 14pt dark blue, and H2 in Arial 13pt bold medium grey.

[quaint eagle]

that's pretty darned good!

doesn't work?

[fast oxide]

Nope, no change

Do I have to apply it somehow?

[quaint eagle]

just run it

[fast oxide]

Yeah, no change.

[quaint eagle]

interesting! can you share your opal? I'd love to take a gander

here's my try https://opal.withgoogle.com/?flow=drive:/1BXjmxeJGAms2mPQcblGtJLbnCpXlm66f&mode=app&shared=true (using your prompt)

[fast oxide]

https://opal.corp.goog/?flow=drive:/11FumLs7rzRfE2GSpbLk5JumveoQLObSW&mode=app&shared=true

[quaint eagle]

oh -- you're a Googler 😄

[fast oxide]

Yes

Your app looks good!

Maybe I need to add Style Webpage to the end of the flow?

How do I delete steps?

[quaint eagle]

select the step, then hit "Delete" on Mac or "Backspace" on Windows

[fast oxide]

Thanks!

[quaint eagle]

yikes! try signing out and signing back in -- does that fix the problem?

[quaint eagle]

just for fun, here's a mini app that generates a blockbuster movie trailer based on a topic you supply
https://opal.withgoogle.com/?flow=drive:/19tDnTsvGv13VMFPoH9xAv9lHxzmTfz4h&mode=app&shared=true

[thorny grotto]

Oh voww

This is so cool

Opal feels like Firebase Studio but with pre built UI and can be used with no dev knowledge

[steady wraith]

I want to be part of Early supporter of Google tools and product? How to apply?

[thorny grotto]

if you mean trusted testers then there's a form you can fill up at labs.google.com at the end of the page

[compact tangle]

Any method to copy the suggestion given. I find the suggestion of modified prompt is helpful but I cannot copy and paste. I have to input it manually.

[ionic prism]

How can i try opal ?

[analog karma]

https://opal.withgoogle.com/landing/

[dense totem]

As long as you're in US, you will have access to the tool

[dense totem]

Thanks for your feedback. Looking into this as we speak

[dense totem]

Can you please share the entire app description text so we try to reproduce the issue?

[pulsar timber]

In reference to his problem, heres what hes talking about i think: you cant copy & paste the suggestion prompt it provides. This is on the 'build your app' interface.
(ignore my prompt; just thought i'd provide an example as i ran into the issue in the past)
See image:

Plus the visuals are a bit glitched for that popup.

[boreal cape]

Hello everyone

[modern badger]

Hi @dense totem Just signed up for the public beta - getting an error when trying to open any of the 'example' featured apps like City CGames or Claymation

[dense totem]

Yes @modern badger , we're aware of the problem and workingo on addressing the bug. Thank you for your patience. I keep you posted. Sorry for the inconveniences.

[modern badger]

No worries - excited to play with this new tool (though still holding out hope for Public Gems at some point) 🙂

[dense totem]

Appreciate it. Can you elaborate on what you refer to by public gems?

[dense totem]

@modern badger can you reload the page and see if the problem goes away?

[modern badger]

Sure, today I can create sharable via link applets in the AI studio / playground and now AI applets via Opal, and I can also create Custom Gems from within Gemini to use personally, share privately or within workspaces. The missing link for me is the ability to share with and explore how others are using custom Gems (similar to Open AI's Custom GPT Store or Poe's Bot Marketplace, Instagram Avatars, etc).

[modern badger]

The issue is resolved now!

[dense totem]

Totally makes sense. This is on our roadmap.

[modern badger]

Bonus points if you call it a Gem Store 😉

[thorny grotto]

I wish there was a Google Stitch Integration with Opal

That way we could get a custom UI for our app as well!

[modern badger]

PS - the use case I'm thinking of here is for SMBs to create marketing gems that they can link to their GMB profiles, among other ideas.

[past glen]

I tried to make a simple chat app, i just prompted, make an app that will chat like a friend, and this is what I got,

can someone help?

Also in the website, this one, that discord link thing

it doesnt give any discord server invite

someone just copy pasted the chanel link there, i was so confused whats wrong, i had to search google labs discord in google and then only i can join here

you cant see a channel in discord if you arent part of the server

May I also ask if there would be a dark mode for this website soon ? sorry I have so many complains, I guess I just expect all google things to be perfect

[past glen]

Ah, same here, I thought my internet was slow 😭 i even restarted my router 2 times, i was so excited

[pulsar timber]

Considering they just released it for public testing, surely its not going to be perfect thats why its good for us to ask questions & suggest changes. I assume over the next few days some of the initial kinks will be worked out. Cheers

[quaint eagle]

📢 apologies everyone, we're having a "Success Disaster": didn't anticipate so much interest in this experiment, so we're working to up our quotas and rate limits. Please hold

[quaint eagle]

oh, def looks like a failure on our part. I am guessing we can't yet build a thing like that. But I love your idea -- will add to the list of things to teach Opal to build.

[quaint eagle]

good to know -- we prob should point to the Discord server directly

[vague coral]

This may seem like an odd question, but is there anything we shouldn't share on social media about Opal besides obvious?

[quaint eagle]

share away -- hopefully, all of our quota bumps will propagate shortly and we'll be ready for new users

[vague coral]

Loving the experience so far! I really think a lot of the future with AI will be about personalizing how you use it like this. Still poking around and figuring out what works and what doesn't.

[steady wraith]

But I am not from United States 🙁 I am from India and want to be part of trusted tester and Early access team

[pliant path]

The design is not optimized for mobile devices.

[exotic tangle]

Are the system prompts for the models within Opal different in any way?

Because sometimes Gemini doesn't exactly follow the instructions... if in any part something happens or it gets derailed, then the output will just crumble.

[spring coral]

this agent has not been working

[steady wraith]

Token number exceeds 🥺

[past glen]

the images load now! thats great!

Look how much better it looks with the dark mode Chrome extension in Google. While you're making a website at Google, since you don't make it in dark mode, don't your eyes hurt? I am curious, my eyes tear up in light mode 😭

the problem with dark mode chrome extension is that it doesnt make the proper dark and some texts i cant read well

[past glen]

will opal be as powerful as n8n automation in coming days using googles whole ecosytem?

[dense totem]

Correct. In public beta, we're mostly focusing on the labtop browser experience. A more first class experience on mobile browser is in our roadmap

[dense totem]

Thanks for the feedback @exotic tangle You see the source of truth prompts that is being sent to Gemini in each step. Can you provide more details on the issue you're describing, with an exmaple ideally?

[dense totem]

Thanks for the feedback. Mind sharing a bit more about your use case, and ideally share your Opal?

[dense totem]

Yes, makes sense. Enabling dark mode is certainly one of roadmap things Opal will eventually have

[dense totem]

We currently support reading from YouTube/Google docs/slides/sheets/etc as well as writing to Google docs/slides/sheet (See the dropdown menu in 'Output' Step on the graph). Yes, Opal will continue to expand tools to support the broader Google ecosystem.

[vague coral]

Any chance of it ever supporting Gemini text diffusion? That would be really fun to stick in an app. Look forward to seeing the mobile expansion! Like a Google play store for Opal apps 😂

[pliant path]

approximately 30 minutes and still ending

[shy crypt]

A couple of things. If the prompt boxes are yellow then that means they are not working, is that correct? Also if you make a prompt box by accident is there a way to deleted it, because I have tried everything and it just lingers.

[quaint eagle]

yes! I really want that too. We want all the fun models, like Gemma as well

[quaint eagle]

2.5 pro can be quite slow. Try switching to a smaller model, like 2.5 flash or even 2.5 lite. I found that they're all pretty good at web page generation

[quaint eagle]

the color of the step denotes its kind:

• yellow = input
• blue = generate
• green = output

to delete a step, click on it to select it then hit "Delete" if you're on a Mac or "Backspace" if you're on Windows

[dense totem]

It seems like you’re using a YouTube video that may be exceeding the 1 million token limit that we currently have. Mind trying a shorter video and see if it works?

[dense totem]

Also, we can look into it in more depth if you could share your Opal please.

[silent lynx]

simple search in step keeps giving me "have been unsuccessful due to exceeding my quota." even using 2 flash...

anything with search web is getting blocked

[dense totem]

And just to confirm, you meant 30 minutes not, 30 seconds correct?

[silent lynx]

"I am unable to complete this request due to exceeding my quota."

[pliant path]

Sorry about removing the Opal, I was waiting for 30 minutes and it wouldn't finish running

[dense totem]

Thanks for reporting this. We’re looking into this as we speak.

[silent lynx]

thanks!

I guess that is going to bring up the request for versions.... since it was working, and I assumed it was me, made a bunch of changes, that I now don't know how to get back to the happy place.

[quaint eagle]

there's a somewhat hidden way to do that: click on the three dots on the top right, and there's a version history there

[pliant path]

https://opal.withgoogle.com/?flow=drive:/1mh5u1WBXuazg8IuQ1SfNem33TZjnnp_M&mode=app&shared=true This opal has the same problem

[elder wigeon]

We have just updated our 2.5 model family search quota to unlimited. Just trying to confirm if you are still hitting that quota issue

[silent lynx]

Yippie ok jumping back in to test it

[silent lynx]

still getting "due to exceeding my quota. I am unable to provide a recommendation at this time."

maybe I jumppped in too soon?

[elder wigeon]

OK can you share your opal with me?

Let me check what's going on 😃

[silent lynx]

https://opal.withgoogle.com/?flow=drive:/12hzY-WlFHwCHjITZpgmglSWCN6TCRz2o&mode=app&shared=true

[elder wigeon]

Thank you!

The error I got seems to be related to access the URL failed. May I ask your input query which triggers the quota issue?

I actually find the issue

[silent lynx]

I have been testing with; Dove body wash, Qtips, coffee drinkers in NYC, Tea drinkers in NJ, back pain treatments.... Earlier today it was doing great, giving good sources for firms that have expertise in say CPG because of Qtips, but the next search on back pain was firms that had experience with medical research... then I hit the quota error

[elder wigeon]

Thanks for reporting!

I will escalate now.

Seems that we are calling 2.0 flash model in some cases.

Will work on the fix

[silent lynx]

cool - when I specify in the card for 2.5 flash I get "I apologize, but I am unable to access external web resources at this time due to a technical issue. "

[elder wigeon]

I also saw that

I think some web scraping failure are there when we are trying to access the link

That will be another bug I am gonna file

[silent lynx]

welcome to the party...

[silent lynx]

YAY thank you! a hidden gem! (well I guess we can't use that phrase anymore... a Gem has a whole new meaning now 🙂

[quaint eagle]

try now

[pliant path]

It's works, thanks

[lavish kernel]

Anybody else having troubling sharing an app from Opal? When I go to share it, the URL always takes me back to the landing page, not the app itself (yes, I have it on public view).

[exotic tangle]

Not that I’ve seen anything specific on Opal, but in general, I use Gemini for almost everything and have done so for the last 6–8 months. Over that time, I’ve noticed that the responses don’t always follow the main prompt as consistently the outputs can change, which happens with pretty much any proprietary LLM. Still, with Opal, I don’t expect it to be shaky or vary so much with every output...

in short i wanna make sure that it will work every single time

in suggestion, if you guys can add a system root prompt to follow the user prompt robustly kindof specially with opal... that would be niceeee

[dense totem]

Thanks for the feedback @lavish kernel . I tried to reproduce this but it works fine for me. Can you take a screen recroding of thie behavior and share with me here?

[floral mountain]

FYI

[hollow quarry]

<hello world!> greetings from a Brazilian in Norway. Has anyone tried embeding opal on a wordpress website? Lot's of love from the vikings land

[silent lynx]

Nice to meet you. I have not tried embeding yet... still getting the same sharing issue as @lavish kernel but I am sure that is going to be the next ASK... "this is great, now get it on the site"

Hey Vitor - My family is in Sao Paulo

[hollow quarry]

I'm intending in use an iframe under an internal site, my idea is to use the books I read during my master's to make lessons.

[silent lynx]

that is a great idea. I am testing this out because we have new students starting in the Fall and historically there is a 200 page binder they get to get themselves oriented with all the resources around campus.... but that is Old, expesive and static... we are playing because if we can give them a link, then it can help them navigate everything from where is the libarary to what clubs play disc golf... and it can link them right to the group

[hollow quarry]

Opal also should connect with notebooklm and use the fximage

[strong slate]

Pls when would the access be available in Nigeria?

[ebon raven]

Dang, nice to see the level of support in this channel from google staff... vs others.

[elder wigeon]

Ooof most likely we got some quota hit. Thanks for flagging

[elder wigeon]

JFYI, iframing might not work for now. Since we do need to nail down how to safely enable embeddings. And iframe is actually... Not very safe 🤫

[hollow quarry]

Thank you for the heads up 🙂

[cerulean basalt]

Hi! I'm experimenting with Opal to create a content generation engine for a client. It's working really well, but a) is there a way to render the final output in text or as a Google doc, rather than as HTML, and b) either way, at the last stage of my flow, I want the user to be able to suggest edits to the created content, but the output renders in a small box that makes it difficult to read the content. Here's a screenshot. Any suggestions would be appreciated.

[silent lynx]

Can't you do that through an output?

[cerulean basalt]

Haven’t tried but I’ve made a lot of progress just by editing the prompts. I’ll make a version that outputs to Google Docs and see how that works - thanks!

[cedar drum]

Can we have an option to export to many formats, or an API. I want to have my app send a message to a remote server to download something to the user's device.

[unreal trout]

Suggestion: it would be great to have Google Calendar as a Tool. This would allow for both: a) using calendar events as context or input(s) into Generate steps b) the ability to write/update Google Calendar as one of the options in the Output steps

Loving Opal btw!

[quaint eagle]

definitely. Adding this as a feature request now

[quaint eagle]

yeah, that sounds really good. Would something like MCP work for this use case?

[quaint eagle]

what browser are you using?

[cerulean basalt]

Chrome

[quaint eagle]

would you mind sharing it? I'd love to troubleshoot the rendering

[cerulean basalt]

I spent some time editing the prompts, simplified the instructions and now it's working pretty well. I'm reluctant to share because it's got a lot of proprietary info

[quaint eagle]

totally understand

[cerulean basalt]

But in case it's helpful here's the before and after:
Before (rendering not working properly): Generate a well-structured HTML webpage with a clean, modern, and responsive layout. The page should feature a clear header with a navigation bar, a main content area for the verified content, and a concise footer. Prioritize user experience with intuitive navigation, visually appealing design, and smooth scrolling. Ensure the layout adapts seamlessly to various screen sizes, providing an optimal viewing experience on desktops, tablets, and mobile devices.
After: Generate a modern, responsive HTML webpage. The page should feature a prominent, fixed header that remains visible during scroll, a scrollable main content area, and a concise footer. The main text content should be left-aligned. If an image is provided, it should be prominently displayed within the main content area. The webpage must adapt seamlessly to various screen sizes, ensuring an optimal viewing experience across desktops, tablets, and mobile devices.

[quaint eagle]

awesome, thanks!

[jade brook]

Just created Inventory AI. How can I integrated this and shoot me high/low %s based on food amount spent? (for a small cafe) If anyone would help me on this, it would be highly appreciated: https://opal.withgoogle.com/?flow=drive:/1bP9NWcKxSMkSJWCsHlIcuyUUzY6fZ99Y&mode=app&shared=true

[oblique crow]

where should we share the apps we are creating on the Opal app? Here?

[oblique crow]

or do we just "publish" from the app?

Is there a way to change the look and feel of the UI of the apps we "prompt into existence" using the Opal app?

[tender slate]

Hello, I've encountered an issue in Opal where searches in languages other than English do not complete.
If a workflow is designed to perform a search query in another language, the search process starts but never finishes and fails to return any results. I specifically experienced this problem with a query in the Czech language.

[dense totem]

Yes, correct. As @silent lynx said (thanks for the response), you can write to Google docs/slides/sheet from he dropdown menu in an Output step. Once Opal is run, you can find the generated Google doc in your Google Drive / Home folder (for the account you logged in with). You can also define the doc's name under the Advanced Settings

[dense totem]

You can click on the 'Share' button, toggle 'Publish your opal' from private to public, and use the URL to share the app (send the URL)

[dense totem]

You can control the UI look and feel of the 'output' generated when you run the app, but changing the UI of the Opal itself is not supported. Can you tell us more about what's your use case?

[dense totem]

Thanks for reporting this. Yes, we're aware of this issue (is related to a decoder bug on our end), and plan to address it next week. Stay tuned (and thanks for your patience)

[dense totem]

Thanks for sharing this @jade brook. I took a look at your Opal app and it seems like it's mostly about capturing trends in your inventory data and offer optimization opportunities. Are you stuck at any particular part of this? Which part do you need help with to get unblocked?

[oblique crow]

Thanks

[oblique crow]

I guess I'm thinking about the app itself... Like when you share the app it's not very personalized... Unless I'm thinking about it incorrectly... For instance, I just made an app that is a simple greeting card generator... I'm still working on making sure it can catch human input errors and letting the end user be able to "view" before completion... Anyway it works pretty well so far .... But the interface, if it were to become a standalone "app" is not nice looking. Is that something I would create in a different app or does Opal become the place where you also create the UI? Or is the "output" that we're making the "UI"? Sorry for the tldr

[red sinew]

Not sure if others are having the same problem, but the app I created generates HTML files, but I'm unable to export (download them). Anyone else run into this problem?

[tender slate]

Thanks

[severe raptor]

any plans to make opal available in EU ?

[plucky sphinx]

same. how can i access it in ID?

[summer zinc]

You can use it in EU there is a way 🙂

[silent lynx]

omg - if I could link my MCP creds as a tool in Opal it would unlock so much....

[burnt meadow]

I'm loving opal so far, just wondering if there be a way to export and embed opal's into other UI's

[runic cedar]

💵 ?

[thorny grotto]

will we always have free Veo 3 on Opal?

[past glen]

Its currently available only in the US, maybe just use a VPN to use it from outside the US

[willow pilot]

Hallo

[hollow quarry]

@quaint eagle since I saw you taking notes on requests. I dont want to ask too much, but could you include fx_image to the image generator and notebooklm to the imputs? 🙂

@silent lynx I've sent you a pm. I'm from Sao Paulo too 🙂

[hollow quarry]

I couldn't find you on linkedin

[quaint eagle]

imagen 4 is the same model that's used in ImageFX, so you should get very similar results

on notebookLM, we absolutely want to add that 👍

[quaint eagle]

@elder wigeon

[tribal terrace]

Hello peeps! I'm impressed with Opal 💖 Honest N∅⊷c∅de congratulations to the team <@&1173646484311384114>
In a managers note, I love the system prompt. It is really like talking to a Junior about "holding its horses and execute the task as told" … please!

[tribal terrace]

I did notice that, Dimitri. Care to explain the rationale?

[quaint eagle]

we think there's a lot of potential in this particular space and we wanted to build it in the open, with you. Start with something and then look for enthusiastic dance partners to drive it to something amazing

[tribal terrace]

I was 'bout to suggest you sharing a video.
My UX bet is that users that choose Opal 💖 are visual learners.
👍🏼

[tribal terrace]

Imagine that! After I saw your Flow lunch, I immediately thought of agencies that do opening sequences for motion pictures.
Here's what I created with your Opal 💖
https://opal.withgoogle.com/?flow=drive:/19tDnTsvGv13VMFPoH9xAv9lHxzmTfz4h&mode=app&results=110jwD-rx7MVqlXuM3_trTHuPqxVP9weq&shared=true

[tribal terrace]

Interesting! I would vote up this 🐞
Meanwhile, may I suggest a hack? I'm trully not being sarcastic, but how does an Opal that transcribes text from a screenshot?

[thorny grotto]

I wish I could add Flow like capabilities in Opal

in our apps

[quaint eagle]

oooh, tell me more

that sounds very interesting

paint a picture for me

[thorny grotto]

suree

I was talking about if we could add frame by frame and camera setting abilities from Google Flow

with Veo 3 and Imagen

I tried prompting the AI but it did something which is not quite functional

[quaint eagle]

that makes sense! basically, add more ability to create Flow-like artifacts in Opal?

[thorny grotto]

yess more functionalities on our apps for video and audio!

and imagen's camera angle adjusting would be cool too besides Veo

[tribal terrace]

Love the vision. Kudos to the <@&1173646484311384114>
Also, users here are throwing so many fast balls ⚾ at you guys, that you are
never
ever
going to clean that backlog, hahaha!
Happy building! 🏗️

[tribal terrace]

Say, may I add a feature to the "Formats & Outputs" suggestions I read somewhere in this thread? We need JSON!
I tried the current buttons, but they just duplicate [remix] an Opal.

My question on the current UI is if this ‹Manual Output› (right screenshot) can be prompted into developing a structured format.
The scenario would be the user building in no-code it's MVP, then taking its approved results to Firebase Studio to build a robust app. Certainly experience coders would choose Vertex AI/ API. My point is **all **builders indistinct of their digital literacy will make OPAL their working "Wizard Of Oz" MVP for UX research.
Thanks!

[dense totem]

Adding this to my list of feature reqeusts. Makes a lot of sense. Thanks @red sinew

[quaint eagle]

huge fan of this

[tribal terrace]

You guys are getting a lotta luv. Kudos! 💖
https://youtu.be/FMTiYeW60Q8?si=QBlFTu_2vc1rj0Go

[thorny grotto]

yesss

Opal is so underrated but hopefully with time more people are getting to know it

[neon pollen]

I would love for Opal to be better able to handle bulk outputs and input requests that's just one idea I have sitting on my hands

Also HAI EVERYBODY hope y'all days are wonderful but I'll 😭die on this hill GOOGLE IMAGEFX, WHISK, AND OPAL to me are what i'd point ppl to understand creativity with AI isn't threatening

[red sinew]

Yeah and it seems to be overloading, because its very hard to use right now.

[quaint eagle]

yeah, we did NOT anticipate this much love
https://www.youtube.com/watch?v=2I91DJZKRxs

[red sinew]

yeah definitely unusable at the moment, too bad I was having a lot of fun. Guess we will have to wait a bit until things stabilize.

[quaint eagle]

btw, where are you seeing errors rn?

[low elbow]

Do you have any plans to release Opal in Brazil?

[quaint eagle]

for now, it's US-only. Obviously, I am hoping for it to be available everywhere!

[low elbow]

Oh, yes. I'd like to know if there's a forecast/date for it to be available in other countries, like Brazil.
Thanks

[quaint eagle]

no dates at the moment

[low elbow]

thank u

[gusty badger]

Hey team!

First off, huge thanks for building this amazing tool — I’m honestly super excited about Opal. Within just a few minutes of trying it, I was able to automate part of one of my branding analysis workflows. That felt like magic. Really hoping this evolves further and makes its way to a global launch soon — it's got so much potential! 🙌

1-That said, I ran into a small issue. After I described the process via prompt and Opal generated the initial workflow, I tried running the app — but one of the steps got stuck in a loading state and the process never completed.

Also, I noticed that the generated workflow compressed a lot of the process I described into just 2–3 steps. I ended up breaking them down manually into smaller sub-steps, and that made the workflow run smoothly.

Here's the link in case it helps to check the logs:

My Opal:
https://opal.withgoogle.com/?flow=drive:/1C1XtZla7AgcuyRqSsXI_KiOnFe5TTXLZ&mode=app&shared=true

2-So my second piece of feedback is related to that:
It would be really helpful if Opal could separate each described step into its own block rather than merging multiple steps into one. Not only does that give me more control, but I’ve also found that when working with agents and workflows, breaking things down into smaller steps increases accuracy and reduces hallucinations — since the model can focus more narrowly on each task.

Thanks again for the great work — looking forward to seeing how Opal evolves! 🚀

[feral void]

Hey loving the product.

However, I am getting the following error message:
Video generation failed: No videos generated: The prompt could not be submitted. This prompt contains sensitive words that violate Google's Responsible AI practices. Try rephrasing your prompt, or contact your Google representative to request allowlisting. Support codes: 58061214
When attempting to create a video from a product URL which is designed to help kids tie their shoes.

[quaint eagle]

yeah, the veo model is overly cautious at the moment. I found that any requests to generate children end up in error

[shy crypt]

Is anyone able to upload a picture from your pc I haven’t been able to do it

[quaint eagle]

do you get an error?

[shy crypt]

It’s weird I hit the plus sign and it gives me an arrow but it keeps moving up like a glitch or something don’t know how to explain

[quaint eagle]

ahhh yes! I've seen that too!

fixing this soonish

are you in "Console" view by any chance?

switching to "Preview" will fix stop the pop up from moving

[shy crypt]

Ok I’ll try that see if it works thanks

[vague coral]

I got a funny error today. It wasn't so much an error as it was a message that said it can't make that. 😂 Great technical explanation of why it couldn't as well. Appreciated that! Even gave a suggestion for an alternative prompt that was similar minus the problem areas.

[vague coral]

Nothing bad 😅 just couldn't write/create the specific API endpoints for accessing Firebase databases. Still made a decent office assistant!

[quaint eagle]

btw, that bug is hilarious. Here's a video of me reproducing it: https://www.youtube.com/watch?v=X_Z1ua_7Mp0

[shy crypt]

It’s private it won’t show

[quaint eagle]

ah yeah, unlisted -- YT will ask to sign in

💖 just want to say thank you so much for playing with Opal. Really helps us see the future.

[simple drift]

Used Opal to create an app for finding local representatives by putting in your zip. Took me a little tweaking to get it right but finally working ok

[hollow quarry]

@quaint eagle I will send you a PM, I think I have a challenge

[wanton saffron]

Does Opal allow you to create and deploy workers?

I know it's no-code, but I just want to know if I can tell it something like "Create me a simple website that says "hello world" to everyone who comes in and behind it has a worker that is actually a discord bot. "
-# And if so, can it be deployed? 24/7

[keen osprey]

That's great.

[feral void]

Can you share this with me? This seems really useful.

[rich escarp]

When will it be available in other countries?

[feral void]

That makes sense, I appreciate the quick response.

[shy crypt]

Ok it worked on preview so just have to tweak it because it started out crazy but ended great

[quaint eagle]

love it!

[quaint eagle]

no workers yet, though we have triggers (run when certain condition is met) on our road map

[opal nexus]

Hey please launch opal in India

[wanton saffron]

Ok thx for the information

[opal nexus]

Hey can you fix this

[quaint eagle]

yup, will be fixed in the next push (today)

[odd widget]

I believe I found a bug. I created a simple app that generates Imagen 4 images. They generate the images in each step but the output does not display them in the preview. It does display when I publish the app.

[quaint eagle]

oh interesting! Could you share the app? You can DM if you don't want to share here

[odd widget]

Sure it's https://opal.withgoogle.com/?flow=drive:/1OWNBiMaxkAMFU69U4cLB5M1Yc0DxwDcD&mode=app&shared=true

[quaint eagle]

love the idea 🤩

[odd widget]

thanks. I also add an update to "suggest an edit", hit the arrow to continue but nothing happens.

[quaint eagle]

re: continuous loop of interaction -- we can't do that yet. Basically, the output step is the end of the flow, and more like "here are the results". Can't loop back to the beginning just yet

here's what I got: https://opal.withgoogle.com/?flow=drive:/1OWNBiMaxkAMFU69U4cLB5M1Yc0DxwDcD&mode=app&results=16pHlcJ0A9dUTWYIcQUfXQ2S5lZp9hijk&shared=true

[odd widget]

When do you think looping may be added? Maybe require a user input prior to the loop?

[odd widget]

It's good but I wanted a selection to do something. I suggested it in the edits but it doesn't do anything. Also your images come back the same everytime I refresh the page.

[quaint eagle]

yeah, it's just one result. Think of the current opal mini apps as one shot tools.

[quaint eagle]

not sure when, but I definitely want it, too

[odd widget]

I'm also interested in integrations with outside open source apps using Fal or Replicate. I think I'll VIBE code on some other stuff until the looping & integration gets introduced. Right now it's pretty limiting for me.

BTW, I remixed what you did & the preview now works. https://opal.withgoogle.com/?flow=drive:/1Jg_nNYGLFSxX_jzzKDWZDyd_KNdI9TN0&mode=app&shared=true

[vague coral]

That's the same message I got! Or similar.

[bleak anchor]

hehe

[dense totem]

Quick follow up here: can you please elaborate on your use case? Curious to learn what's your use case, and how you're planning to use the downloded HTML?

[dense totem]

Can you tell me a bit more about your use case that requires looping?

We're looking into it as part of our roadmap, but I'd love to understand the nunaces to make sure we're aligned

[dense totem]

Just to make sure we're on the same page, are you referring to the issue of UIs overlapping, or the fact that your intent resulted in an error?

[odd widget]

In general, looping will be essential because I believe most apps will need to start over once you've done a task at some point. In my case, a user will select the image they like and the next logical step was they would select between two images again.

[cedar drum]

Yes. As long as it can retrieve data from somewher (e.g. Google Drive)

[opal nexus]

yes bro

[opal nexus]

Yes I am referring to the part of Ui overlapping.

[stray swan]

Will there eventually be a away to export the workflow from Opal > Agentspace, or JSON? I realize that's not the intent of the product, but for someone using the ecosystem I could see a lot of value there. Wish list(not aligned with product direction): output google ADK code for the sequential flow 😄

[quaint eagle]

yes. We already have the ability to export, though it's a bit hidden: just select all (Cmd + A on Mac, Ctrl + A on Windows) steps in the editor, then copy (Cmd/Ctrl + C), then paste in your favorite editor

[thorny grotto]

Hopefully when computer use (Agentic capabilities like Project Mariner) gets generally available, would love to create agentic apps with Opal 🙂

[opal nexus]

Getting this error

[quaint eagle]

uh oh. How did you arrive at this error? Could you give me a bit more details?

[opal nexus]

Used an prompt "You are an expert in SEO and content strategy, specifically tasked with generating highly effective research queries. Your goal is to formulate a detailed and specific query for the identifying long-tail keywords and blog writing insights for a given country and topic. The query should be optimized for comprehensive research, aiming to uncover nuanced and actionable information for blog content creation." Opal created an agent but when i started to give the prompt to research it throws an error

[dense totem]

Thanks for clarifying. Seems like in this case, you want to be able to re-run / loop the entire Opal app? vs looping a sub-graph, correct?

[dense totem]

We already think of Opal apps as 'agentic app', as we offer agentic components such as deep research, or an agentic step called 'plan and execute' (under Generate) step, but would love to hear. Tell me more about what type of agents you're thinking of?

[thorny grotto]

I'm thinking about computer-use Agents

I can give some examples of apps I could ask it to build

Here's an example:

Build an AI app that could check my today's Grocery list on Instacart and then make it use Gemini for producing three recipe ideas based on the incredients and then write the recipe and also make a video of each using Veo

or maybe something like:

An AI app that could check my latest movie history on Google TV and then give me suggestions based on that or create a short story using Gemini combining the characters and creating a short movie like trailer

Just some of my wild ideas lol. Maybe not too accurate 😆

[thorny grotto]

In short besides deep research, I would love to see something like Gemini's upcoming feature called "Agent Mode" (announced at Google I/O) in Opal

also can you tell me a bit about what you meant by "plan and execute (under Generate)"

[odd widget]

Correct. This would also help with the score. Once the loop has been made, the score gets updated by one.

[stray swan]

I 100% Agree! From a product perspective, Opal looks like an easy way to get customers to test an idea for a workflow, then quickly export that JSON to agentspace as a starting point. I imagine they're just in the data collection phase before we see anything like that. These JSON flows would also lend well to boilerplate ADK setups, they'd be crazy not to take advantage of that IMO. Opal creates a streamlined path to agentspace adoption BEFORE the customer has per-seat spend tied to agentspace, while being able to vet the workflow idea. I've really enjoyed Opal so far! Very fun!

[thorny grotto]

yesss exactly my thought when I tried out opal

It helps me to get like a mindmap on the project that I'm about to work on

[fathom wren]

I am a teacher and used Opal to create a test feedback and practice questions after inputing in test results from a class. It works great but am wondering how I get the results to students. I have it setup that the output would be each students' results would be on a google doc but that doesn't seem to be happening. Any suggestions would be helpful.
Also I can't seem to change the sharing permissions of my Opal so I am a little hesitant to make the Opal public.

[red sinew]

It creates a landing page based on preferences, it should also output a css file.

[viral prawn]

Anyone know if I can download the app created in opal and save the source code in github? I want to plugin the maps but opal wont allow that.

[wanton blade]

When you say it doesn't seem to be happening, are you seeing any errors at all?

[jaunty sleet]

Is there any basic scripting? For example, can I have the output filename be input filename + ".json". Also, I would love a REST API.

[wise goblet]

Hi all — I'm trying to figure out how to create a flow in Opal that asks for additional user input if the model still needs additional information. The flows I can create are linear and don't allow for the model to make decisions like that ("if I need more information, I'll ask for additional user input"). Has anyone run into this or have any ideas? Thanks!

[fathom wren]

I see the outcome but not as Google docs. In my Google drive Opal folder there is what I think is a JSON file that is titled results. I am hoping that each student’s feedback would be on its own Google doc. If that is happening, I am not sure where I would find them.
I am not seeing an error messages

[dense totem]

This is very interesting. Thanks for sharing @wise goblet . Can you tell me more about the exact use case you're working on?

[rich rain]

How do we download a blog created on Opal?

[quaint eagle]

not at the moment, though it's on the roadmap. For now, the only way to save the output is by using "Share output" button:

[quaint eagle]

no scripting -- and my hope is that we'll never need it. Looking forward, the way we're thinking about it is that that the user will use natural language to describe what they want and if this writing code to process data, Gemini will do that job -- perhaps as part of a special step.

I see the outcome but not as Google docs

How do we download a blog created on

[quaint eagle]

don't know if this is what you're looking for, but here's a pattern of how to use "Review with user" to create a chat loop: https://opal.withgoogle.com/?flow=drive:/1-OTBAIWLyGeDMJ2IWEJ286ZYZR8P-FcO&mode=app&shared=true

[lyric rampart]

Does anyone know how to insert Javascript into a module?

[slate crane]

Hello, for some reason it says Opal is not available in my country. Is there a way to come around it?

[dense totem]

Opal is currently only available in US. We plan to expand it globally in the future but don't have any firm dates yet.

[dense totem]

It's not currently supported, but can you tell me more about your use case and what you're trying to accomplish?

[slate crane]

Thank you, I messaged you

[vestal spoke]

Hello all, I'm working on an Opal project and would like to use triggers to and from google sheets. My Gemini is telling me to use the trigger function in Opal but I don't see one available to me in my instance. Is this a hallucination from Gemini?

[slate crane]

I am trying to build an app

[granite wren]

Is access to opal invite only? I’m in Canada and got message it’s not available in my country?

[broken prairie]

There's no way to run an opal app automatically on a cron is there? Additionally, is there a way to trigger it via curl or not having to manually click start every time?

[paper patio]

I was going to say the same thing, was excited to try Opal and not available, I'm from Portugal

[quaint eagle]

yeah, it's a US-only launch for now 😢

[quaint eagle]

triggers are on the roadmap

[merry cedar]

Is there a user guide or some kind of instructions to be able to properly use what opal can do? I’ve begun building but there are some things that need a bit of help.

[woven yew]

Could OpaL be integrated with some sort of API (vertex AI?)

[lucid harness]

Guys.

I am trying Opal for the first time.

[swift furnace]

<@&1173646484311384114> Are there any ETA to roll-out opal outside of US?

[fickle cedar]

possible with vpn?

[slate jungle]

Is there a block limit for an app?

[wanton blade]

Welcome! How are you finding it?

[lucid harness]

It's...GENIUS!

[lucid harness]

In the USA, it's not.

I'm from Serbia, so I'm using a VPN.

Using Opal is like Game Builder Garage but for AI apps.

[royal wharf]

Posting this here too in case general was the wrong place hehe

Hi guys, I made an app with optional inputs and must have made a mistake, when I left the input empty the app started 'waiting for user input' and has been that way for "1600 seconds" so far. How do I stop it? I clicked what looked like 🔃 thinking it would reset the app, as well as editing the app, nothing seems to interrupt it. I thought I might just copy everything into a new app and delete the old one, but I can't recover the input I used to generate the app (nor can I make it add a node containing the original generating prompt, it didn't understand.)
If it helps, the idea is you input a general prompt (text) and a list of 'aspects', each aspect can have text and/or image examples, e.g. 'styles', 'media', etc. I wanted it to fill in the blanks with content aligned to the perceived intent before generating an image incorporating all of the aspects.

[lucid harness]

I used Opal to make a Wikipedia article generation app for non-existent stuff.

Here it is!

YEEHAW!

That was fun.

[opal nexus]

Use vpn we have no choice till now.

[young pier]

Unfortunately it doesn't seem to be available in UK. when will it be released in UK?

[opal nexus]

There is no ETA from team. They are working on this I am also using via VPN

[lucid harness]

How do I move a step to the prompt on Opal?

I'm trying to move one.

NVM, I found a way.

https://opal.withgoogle.com/?flow=drive:/1QGE3cgwC0zem50Qoo8hHf-SVZS4U5rXu&mode=app&shared=true

[lucid harness]

I have a question.

Is Opal unlimited?

For me, I think so because it's experimental.

[wanton blade]

There is a "limited free quota" on some of the models.

[tribal terrace]

Elara Vance (as Lena Thorne) is still making waves, hahaha!

[viral prawn]

Thanks i just reviewed share output result and it makes sense, i can convert into python app as weekend challenge 😉

[lucid harness]

Like Veo 2.

[lucid harness]

Where's the full article?

I tried it and it's not working.

It's barely responding.

It's slow.

[past glen]

hello, may i ask? why dont you add gemini in google translate?

google translate is always wrong in this era of AI

i hope google team fixes it

its very impractical to go to gemini every time and ask in chat

google translate has been around there for years, why dont u guys add AI in it?

you have already added AI in mails, search, youtube and so many new AI tools

every time i use google translate, let it be whatever languae it gives a trash response

i always have to use gemini or AI studio to get the correct translation

please guys add gemini in google translateeeee

🙏

also you have added voice dubbing in youtube which translates any language to english, can you guys add those real languges gemini 2.5 pro uses? it would be greatt

[bleak anchor]

hey Kat! this would be great to share with our Gemini friends (which isn't under Labs) - feel free to join their server discord.gg/gemini (FWIW, i've passed this feedback along too!)

[past glen]

omg thank you so much! google is a life saver, and basically i am using gemini to learn new languages using gemini

its so good at explaining anything and beats other large languge models like open ai, claude and anything, i even forgot when was the last time i used open ai or claude

[past glen]

sorry to ping you again, may i ask is there a google chrome discord server ^^? thank uuu

[bleak anchor]

so glad to hear you're loving Gemini! I don't know of an official Chrome server, unfortunately!

[royal wharf]

@bleak anchor I seriously love Opal already but when my tab or browser crashes I lose all my progress - please make failure recoverable, if not the process that was running then at least the inputs up to that point. You could also make it so you can run an app with an existing input set so you only have to provide overrides to the inputs you want to change, and/or make it so you can backtrack to a node to change your inputs.

(I used bold so you could skim but in retrospect it looks kind of rude, I didn't mean to 'shout' lol)

[past glen]

ah so theres no autosave like google docs for example?

[royal wharf]

From what I can tell (and I'm new to this) the app design autosaves but not the inputs or progress into a specific run.

[trail mica]

Has anyone worked out "Save to Google Sheet" as an output. Always seems to fail. I tried a pre-step that was convert to CSV format before thinking that might help passing into it -but still nothing.

[stark falcon]

Loving Opal. Its helping me understand how automation works

[lament hearth]

how would I try out opal?

[bleak anchor]

at https://opal.withgoogle.com/ 🙂

[quaint eagle]

Has anyone worked out "Save to Google

[dense totem]

At the moment, the video posted on the product home page is the best place to start, but we plan to post additional content here and as product documentation in near future. If you tell me a couple of areas that you need help with, I"ll be happy to answer any questios, or prioritize those

[dense totem]

Just curious, can you elaborate a bit more on what you mea by "move a step to the prompt"?

[dense totem]

Yes, the changes that you make to the graph or prompts are auto saved. However, currently, the inputs that you provide to the app is not saved. Also if the app doesn't complete execution, it doesn't save the state. Both of the latter features are on our roadmap to address

[royal wharf]

Great, thanks for letting me know!

[merry cedar]

I’m trying to figure out how to get the output to Sheets to work.
Right at the start, instead of prompting, one option is to add a module (I think that’s what they call it) but I couldn’t figure that part out.
Got a couple working from prompts though.
Amazing to watch how the pieces start working together.

[merry cedar]

I would be selling Opal as something along the lines of “AI Automations + Outputs = Opal” or “VibeCode Automations” or “ Vibe-Automation Mini-apps”

[royal wharf]

Hmm, there's a copy button here, is there just no way to use this "Model input"? Also, will the bot hang like this if moderation is triggered at some point during a run? I haven't encountered an official "error" yet so I'm not sure if it would report the issue.

[thorny grotto]

Hi just wanted to know if the video generation feature with Veo on Opal uses Veo 3 or Veo 3 fast! 🙂

[neon pollen]

Well that's good news 🙏🏾

[frozen tundra]

Does anyone know if, like, when I create an app on Opal, I can actually save it and use it as a regular app without having to go back to the Opal website every time?

[dense totem]

Your app is always avaiable as a URL, hosted by Opal. You can also publish your app and make it available to others by copy and share the URL that is provided when you click on 'Publish'. The app is hosted within the Opal environment, meaning you don't always have to start from the Opal home page / website

[dense totem]

Tell me a bit more about your use case, and what you mean precisly by a 'regular app'.

[frozen tundra]

like a Gemini app for mobile for example @dense totem

[dense totem]

We have limited quota on Veo model family, but to provide the best user experience and minimize interruptions due to the lack of quota, Opal uses Veo3 as default, but then falls back to Veo3 fast and Veo2 as the max quota for each model is hit.

[thorny grotto]

Would you mind telling me the limits of the Veo 3 (default set model)?

[fluid onyx]

Hey Guys, i just found opal, is this worth having as an agent? cna this help with building make.com workflows?

[midnight crypt]

opal vs idx vs cursor?

• are opal apps/chats used for training? are they private?

[pulsar beacon]

When Opal realease in VietNam

Can't Wait to test it

[silver drift]

Hi, I was wondering how to use input correctly. I wanted to have the gemini prompt for more information after the initial input, but not sure how to do that

are there any documents for how to use?

[opal nexus]

https://youtu.be/E0hrcDO3Noc Watch this video by google

[tulip ravine]

who is this girl🤨

[opal nexus]

google employee

🤔

[tulip ravine]

how to make cold my pc

[velvet mantle]

May you elaborate on what the Plan and Execute mode does and when to use it? Thank you so much

[south dew]

I was looking to call one of theses Opal apps from another app, or to include into my app. If I look at the console it looks very much like something built with ADK. TIA

[limber pawn]

Seems like the LLM generated app can have hard-to-use UI

[wanton ocean]

when it will be available in Brazil

[wet hamlet]

Hi Opal team, is there any way to get beta access outside of the US?

[royal magnet]

where do one send bug opal reports

[nova umbra]

I specifically came for opal. It shares a lot of dna with my app I’m building.

[quaint eagle]

May you elaborate on what the Plan and

[quaint eagle]

where do one send bug opal reports

I was looking to call one of theses Opal

📢 Opal is US-only for now and we don't have any data on when/where it will become available next. We're working on that though 🚄

Seems like the LLM generated app can

[civic dagger]

Hi Opal Team! I’m currently in India waiting for open access, but had a quick question in the meantime. Are tools or MCP integrations currently supported in Opal, or are there plans to add them in the future? I know many folks who could benefit — for example, automating preliminary website security testing (running common vulnerability tools, collecting logs, and doing basic analysis). These tasks are usually done manually in traditional web testing workflows, and a flow-based AI system like Opal could really streamline the process. Would love to know if functionality like this is on the roadmap!

[lusty drift]

When available for EU???

[lusty drift]

Maybe Q4 or why is do you have data?

[mortal ivy]

Does anyone has issues generating Video today? With opal I'm stuck on video generating on the app I'm testing

[mortal creek]

hey team. when will the UK get access?

[lusty drift]

Not this year

[rustic nebula]

Hello, I found that when using the OPAL Playlist generator that Gemini hallucinates youtube video Ids. This is exactly the same problem I had 6 months ago with a custom api call. I gotta think you the Google team tested their demo application. Is it just me account that it hallucinates for? I tried and tried to get it to work without success and now I see your application has the same issue. Please advise.

[quaint eagle]

Hi Opal Team! I’m currently in India

Does anyone has issues generating Video

Hello, I found that when using the OPAL

[meager reef]

Hello Opal Team, how do I get access to Opal?

[bleak anchor]

you can try Opal at https://opal.withgoogle.com/ 🙂 please note that Opal is not available everywhere just yet!

[meager reef]

Thanks for the information!

[trail mica]

Does Google retain any of the data from attached files (either via prompting or workflow), and do they train on outputs from Opal while in its experimental phase?

[cedar drum]

Great question

[quaint eagle]

We are doing neither.

[dense crest]

how come google opal is not avalible in uk?

[heady imp]

hi, how do I access in San Francisco?

[cedar drum]

Sometimes the status bar at the bottom of the preview panel in the editor thing gets in the way of the content being shown. Also, sometimes when I switch to the console (I think when there's an error) the status bar stays there and gets in the way.

[cedar drum]

Wouldn't you just go to the link? It should work fine

[cedar drum]

Only when it is awaiting input

[heady imp]

Nope. It asks me to sign in then says "We are sorry, but you do not have access to Early Access Apps. Please log in to your Admin Console to enable it. "

[cedar drum]

Is the account that you are using managed by anyone (e.g. Google Workspace, Cloud Identity ...)?

[heady imp]

Yes, it's managed by Google Workspace, and I'm the admin, yet I don't see anywhere to enable access.

[cedar drum]

How an Admin Can Fix This
Sign in to the Google Admin console: The administrator must sign in using their admin account.

Navigate to Early Access Apps: Go to Menu > Apps > Additional Google Services, then scroll to and select Early Access Apps.

Enable the Service: On the "Settings for Early Access Apps" page, click Service status. The admin can then choose to turn the service On for everyone or for specific organizational units.

Allow Core Data Access: It may also be necessary to go to Core Data Access Permissions on the same page and check the box for "Allow users at your organization to access Google Workspace and Customer Data using Early Access apps."

Save Changes: After making the selections, the admin needs to save the changes.

From Gemini

[heady imp]

Worked! Thanks.

[cedar drum]

Thank Gemini.

But you're welcome

Is the veo generation using veo 3? It's been generating with sound for me.

And I don't have any subscriptions

If it is, please don't change it.

[quaint eagle]

Is the veo generation using veo 3? It's

[wet hamlet]

Noted with thanks. Looking forward to getting access in the future 🙂

[cedar drum]

Can we have an option in the Manual Layout to input HTML? Is that already how it is?

[quaint eagle]

Can we have an option in the Manual

[dense totem]

Sometimes the status bar at the bottom

[cedar drum]

Dragging a file into an input doesn't work

It just says 'Unable to load data'

[wicked marsh]

When EU?

[dense totem]

It just says 'Unable to load data'

When EU?

[tidal grove]

Hello, I am new to OPAL. Can OPAL create frontend code with workflow? Thanks

[dense totem]

Hello, I am new to OPAL. Can OPAL create

[south dew]

When I am building a workflow,how do I remove a workflow element. I see no way to delete. Only way which is not optimail is to go back in history and set it back in time. what am I missing. Related how to I remove a connection? TIA

[quaint eagle]

📢 Opal just got an update. In this release:

• better error messages (instead of a wall of text, we try to tell you exactly why the run failed)
• richer console messages for video/audio/image generation. You can now see inputs/outputs just like for text generation
• other bug fixes

[ashen ravine]

Anytime I try to use opal this is what I get repeatedly

[visual kestrel]

Could anyone share some example projects they have made with Opal - I want to see some of its capabilities

[dense totem]

Could anyone share some example projects

[robust socket]

Apologies for the question that seems a bit silly, but how do I delete a node (e.g. Input, Generate, output, asset) if I no longer want it on the canvas? Sorry if I'm missing something incredibly obvious in the UI 😬

[inner frost]

Guys im new to opel can anyone explain what exacly is the app it creates ? Is it like automation or somthing?

[lyric sigil]

I want to ask how to solve this problem when I choose to create a new page after logging in.

[shy crypt]

I had the same problem all you do is click the node then backspace obvious but not really lol

[robust socket]

😅 omg, can't believe I hadn't tried that...thanks so much!

[acoustic storm]

I select the node, and in the "Suggest an edit" box I type "delete the selected node" - I think the UI is missing this capability.

[acoustic storm]

Aha! Just read the backspace trick!

[mortal oak]

Congrats for the app vision but limited for automate like post on social network

[limber pawn]

Is there a plan to address the cross site iframe issue in the generated app?

[quaint eagle]

Is there a plan to address the cross

Congrats for the app vision but limited

[sullen nimbus]

When it will be available to india

[forest pecan]

I have a workflow like this, but Get Webpage seems like don't work right now, is it right?

[silent lynx]

Output to Google Doc is not working? I have a simple output to a google doc and it makes a good page, but never makes the doc?

[dense totem]

Would you mind sharing your Opal please @silent lynx so we can take a deeper look?

I have a workflow like this, but Get

[silent lynx]

https://opal.withgoogle.com/?flow=drive:/1CpfapErZyMG5i-M_MtxU81GMjeMEUgcc&mode=app&shared=true

[odd ferry]

hey individuals! Does opal have restriction on the type of app I can produce?

[silent lynx]

what do you mean, there are things it can do and things it can't do? Make a mobile app that is listed in the app stores = no

[odd ferry]

I want to make an app that performs checks on a website like WHOIS and ns lookup.

[odd ferry]

Redhat stuff

[silent lynx]

sounds interesting - I wonder how specific you can get in "search" and url....

[wraith viper]

is it possible to connect an opal app with another opal app in the same account? or do you have to build a monolith?

[odd ferry]

What do you mean? The domain itself will be attached to a IP. The domain can tell you what IP registration details, address and contact info. I want to build an app that can easily lookup that info and also give you the details on what versions are being used for operating that site.

[quaint eagle]

https://opal.withgoogle.com/?flow=drive

What do you mean? The domain itself will

[granite flame]

Hey Guy's I'm new to opel can anyone explain what exacly is the app it creates ? Is it like automation or somthing?👀

[rain flint]

Can I see the finished opal designs?

[blissful bear]

hello, hope everyone is doing great. I am new to Opal. I have 2 queries:

a) i have google sheet as an input - can we update the same sheet ?
in the generated instructions I am seeing
<not_callable_tool>update_asset</not_callable_tool> the Reviews asset by changing the 'Used' column for the selected review to 'yes'.

b) i am generating a video and want to save it to google drive but dont see an option. If i save it to google sheet its just saves some timestamps and text about happening in the video.

[whole shale]

Hi All, I was wondering when Opal will be available at my region. Is there any plan for that? I live in Turkey and longing for this right now.

[lyric sigil]

Hello, if the video shows insufficient balance, where can I top up my account?🙏

[wary vigil]

gemini-embedding-001

Can this model called in opal and performed tasks on texts?

[twin marsh]

@quaint eagle hello

Hello

[remote meadow]

Hi

Any plans to enable the use of opal headless? Like an api for Notion, for example?

[lyric sigil]

@dense totem 😫

[lyric sigil]

this one

[quaint eagle]

gemini-embedding-001

Any plans to enable the use of opal

[quaint eagle]

📢 quick update! It's been pretty obvious that deleting a step wasn't obvious. So we made it a little bit more visible: if you right-click on the step, you now get a liiiitle menu that allows you to delete it

[hearty forge]

Same question here 🤷🏼‍♂️

[untold spruce]

Want to have Opal perform looping, for example, if a video script is divided into 5 parts, and you want Veo3 to execute 5 times to produce 5 segments of the video

[gusty badger]

Hi Opal team,

Currently, Opal allows creating linear workflows, which work well for straightforward processes. However, there’s no way to handle conditional branches or exceptions.

For example, I’m building an app to solve complex problems. The first step should be to check whether the user’s input is actually complex. If it’s not, the app should reject the input and guide the user accordingly. Right now, there’s no built-in way to implement this logic.

Request: Please add a Condition component that can evaluate criteria and branch the workflow accordingly (e.g., IF/ELSE logic). This would make Opal much more versatile for real-world scenarios where decision-making is required.

Thanks for considering this!

[wanton blade]

Hello everyone! We just wanted to give you an update on the changes we've made to the Opal codebase this week. Some of these are still Work In Progress, but we wanted to give you a little heads up all the same!

✨ [Coming soon!] Connecting to More Tools (Model Context Protocol Integration): We've laid the groundwork for Opal to connect with other applications and services in a more consistent way. This will make it easier to share information and work with different tools from within Opal. (Note: this is not yet available, we've just started the groundwork!)

🛠️ **A Smoother Visual Experience: **We've made several improvements to how you interact with Opal. Now, scrolling works better in our activity log, and you can right-click to delete nodes in the editor (top feature request!). We've also fixed issues where different visual elements would overlap incorrectly, and we've improved the clarity of visual markers on repeating tasks.

🚨 **Clearer Error Messages: **We've made our error messages more helpful. Now, when something goes wrong, you'll get more detailed and understandable explanations about the cause of the problem, including issues related to safety filters or missing information.

☁️ Faster Access to Google Drive & Galleries: We've made our galleries load faster with a new Cached API. This means you'll see your lists of items load more quickly. We've also made some behind-the-scenes improvements to how we retrieve your Google Drive files.

🔒 A More Secure Experience: We've improved the sign-in process to be more reliable, especially when it comes to checking permissions and handling location-based access rules.

🤖 Better and Faster AI Features: As part of this update, we are now using the newer and faster Gemini 2.5 Flash model by default, resulting in quicker and more efficient responses.

[delicate viper]

why is it always limited for US users

[blissful bear]

Hi, I am not able to save generated content via the workflow. e.g. image or video to either docs, sheet or slides 🙁

[quaint eagle]

Hi, I am not able to save generated

[oak bay]

Called
"The drone stocks are going to the moon"

[deep flame]

app for making videos like veo3

[autumn jetty]

When will this be available in South Africa?

[gentle dove]

hii, what is the file format for opal files in google drive

[wind geode]

When will opal be available in India?

[eternal heath]

Hey everyone, do you know when Opal will be available in France?

[eternal heath]

Honestly I'm looking for the same info for France 😅

[shy dragon]

+1

[gentle dove]

Hii, i created gemini gem / ai studio persona prompt and json schema to generate opal workflows instantly as json objects. PLEASE help me what is correct file format so i can upload workflows to google drive opal folder

[quaint eagle]

Hii, i created gemini gem / ai studio

[glass sage]

When Opal will be available in Romania?

[wanton blade]

It is JSON. If you select the entire contents of your Opal app and copy it to your clipboard with Cmd/Ctrl + c you can paste it into a text editor and have a look around it.

[supple marten]

Vietnamese user expect Opal will be avaiable here soon

[bleak anchor]

fwd from @stoic cedar!

[quaint eagle]

Thread

[forest pecan]

how to copy the output?

i can copy model input, but how to copy output?

[lusty vessel]

I’m getting an error that reads “Gemini failed to generate result due to STOP” while trying to apply a filter step on a website data fetching app. How can I figure out what’s causing the error?

[quaint eagle]

I’m getting an error that reads “Gemini

[cosmic ravine]

I just made my first Opal app, inspired by the Opal team’s “Claymation” app 😄
Nothing groundbreaking , just a light, fun little app to play around with

Photo-to-Memoji
https://opal.withgoogle.com/?flow=drive:/1lfHfuAVuz9u3muQIC_thsvDZ0mVUNsCs&mode=app&shared=true

[quaint eagle]

I just made my first Opal app, inspired

[mental oar]

When is opal gonna be available in all regions it is sad for some of us who do not have a chance to use the platform

[tawdry mason]

When will it be accessible everywhere

[ebon raven]

Does anyone know where there is a list of publically shared Opal apps? User-created.

[shy crypt]

Does it work good

[chrome elk]

It’s an ASMR video prompt generator, specifically for styles like cutting fruits made of different materials. But it also has several other features. There’s a mental trigger section, a stimulation section, so there’s a part that adds only stimulation to the video and another that adds mental triggers to keep the viewer engaged.

[chrome elk]

This part is only about mental triggers and stimulation.

So this is the result, but it’s still in testing.

I’m kind of producing a study on prompts and how they behave when receiving different types of command orders.

[chrome elk]

So this setup has exclusive parts for adding stimuli and triggers to characters made of that material.

[shy crypt]

Its why I asked, you have so many agents pretty much refining the output what I gathered actually came out great

[chrome elk]

Yes, in total there are three: Lúcia, Tatiane, and Vittor.

[dreamy fulcrum]

I just made a 2D game assets generator. but its not perfect. Sharing the link if someone can tune it and add proper prompts that can actually generate all the assets with proper animations will be mind blowing.

https://opal.withgoogle.com/?flow=drive:/1dJy0-iVXJfF4-dI07HiWzU203c-RMDjp&mode=app&shared=true

[chrome elk]

Hello, can I duplicate it and make some updates to it? I had an idea for a sequence and custom agents.

[chrome elk]

The idea is very interesting, and adding custom agents to it is even better.

I’m going to build a setup from scratch with agents now and share it here with you, it will generate reflection and motivation videos.

Or does anyone have a better idea?

[dreamy fulcrum]

Absolutely feel free to use it. I am having trouble while implementing this. If this can happen, it can solve major 2d game dev problem

[dreamy fulcrum]

Absolutely sure sure

[chrome elk]

Ok then I’ll make a copy because I know I’m going to make a lot of changes 😄😄😄

[chrome elk]

I’m Brazilian and here I need to use a VPN to access it, so unfortunately I’m limited in the tool, just explaining in case I don’t log in or can’t do something.

[dreamy fulcrum]

Yep go ahead... all I want is to build a 2d game 😎😎

[chrome elk]

The tool hasn’t launched here yet so I need to use a VPN and that limits me a lot, so I’m just explaining the situation.

I asked to be your friend so we can keep talking about the changes and updates ok

Just explaining in case you felt suspicious or uncomfortable

[dreamy fulcrum]

It's fine. And yes same.... And honestly it's a great idea. We can collaborate together on this. A very handy tool indeed

I have been trying different ai tools to generate 2d assets, but the issues comes in animations

[chrome elk]

Yes yes, in a single setup I put 6 different agents where basically before finalizing they have a meeting to discuss and choose the best option that each one who is already a specialist created

Also discussing whether there were enough mental triggers and styles in the prompts

[dreamy fulcrum]

Hmm really interesting.... 👌👌

[chrome elk]

One of the agents is responsible for mental triggers and for subtly embedding mental stimulation in the prompts, and she is entirely based on knowledge from scientific and academic articles on the subject

It will show in real time which part it’s on and what it’s doing

[bright pecan]

I created my own Veo 3 but the audio is very bad. Can someone help me please ?

[dreamy fulcrum]

A very interesting one indeed.... looks promising....

[vague coral]

Can we remix our own after we've made it public? I tried at the beginning and got an error. Haven't tried since.

[cedar drum]

Who came up with this prompt?
You are working as part of an AI system, so no chit-chat and no explaining what you're doing and why.
DO NOT start with "Okay", or "Alright" or any preambles. Just the output, please.

[chrome elk]

If you want to help I can teach you just message me privately

[chrome elk]

I did.

[cedar drum]

How did you get it into Opal?

[chrome elk]

I tested and built it.

[cedar drum]

Opal?

[chrome elk]

Yes yes

[cedar drum]

Good Job

[chrome elk]

Thank you very much if you need help I’m available my friend

[cedar drum]

When are we getting the feature to allow us to directly put HTML/Javascript into the Manual Layout block? My only/current workaround is to give it to the AI to output.

[chrome elk]

I do use AI, my main focus is on the quality of the agents’ prompts and on the result.

[cedar drum]

It would be much easier if I could just put the code in directly.

[chrome elk]

Just to be clear, even though I have a degree in software engineering and have worked for years as a Java programmer, yes, I am using artificial intelligence, mainly Grok and ChatGPT.

[chrome elk]

I still haven’t been able to understand this difficulty of yours, could you show it to me?

[quaint eagle]

Who came up with this prompt?

[cedar drum]

https://opal.withgoogle.com/?flow=drive:/1EtQioV2Inz-rQjBrl6Qf4PRahtXEnKQC&mode=app&shared=true

[chrome elk]

Open it up, I didn’t understand.

[chrome elk]

I still don’t understand your problem or question.

[cedar drum]

In the output, Render HTML, it would be easier if I didn't have to prompt the AI and could just put the HTML in directly.

[chrome elk]

But that depends on each person — the time you spend just doing this part, which might take you hours or minutes, couldn’t you be improving the structure of the scheme or, better yet, creating more agents?

[chrome elk]

But that depends on each person the time you spend just doing this part, which might take you hours or minutes, couldn’t you be improving the structure of the scheme or, better yet, creating more agents?

[chrome elk]

Are you talking about this part?

[cedar drum]

That's not my project. Ths is:
https://opal.withgoogle.com/?flow=drive:/1EtQioV2Inz-rQjBrl6Qf4PRahtXEnKQC&mode=app&shared=true

[echo marlin]

hello guys, I've had that error several times in a row, do you guys have any idea why this problem occurs ? and is there any solution ? thanks 😁

[chrome elk]

It’s usually an error in the search part or in the part that filters and processes the result.

[chrome elk]

Breaking down your setup, it has several variable parts, so check if part 1 is only doing part 1, and so on

[chrome elk]

Always make sure that each part only does its own part, ensuring that one part does its job and is not doing another part’s job.

[quaint eagle]

hello guys, I've had that error several

[chrome elk]

Hello, my friend, how are you? Were you able to solve your problem?

[regal crane]

does anyone know how to fix this? I'm trying to generate a video from a reference image, there's absolutely NO minors on my image, as you can see: https://imgur.com/a/1J0Y07q. What could be triggering this?

[chrome elk]

Try renaming the scheme name and saving it.

Share the process of a new setup I’m developing.

I’m still creating the structure and base of the agents there will be 6 in this one. Do you want to choose the names? Haha

[agile radish]

Keeping momentum...

[chrome elk]

The result still isn’t good yet.

[quaint eagle]

super-impressive though

[chrome elk]

It’s just that I feel it’s still not that good.

[chrome elk]

😅😅Suggestions?

[modest plume]

Hi can anyone suggest the best way to learn opal use?

[digital falcon]

hey guys, i am patiently waiting for opal release in my region. Any updates for EMEA? If i could even get a preview version i could get the learning curve out the way so i can onboard the rest of my company. Thanks

[wanton blade]

Yes, you should be able to find an overflow menu in the project listing with both "Delete" and "Duplicate" options. Duplicating is Remixing just by another name.

[vague coral]

Thank you! I don't think my idea worked quite how I wanted, but I will have to try it for another. I was trying to make a "Computer Vision Dataset Maker" where you just have to upload the images and it resizes, normalizes, and splits them into train/test. I got the tool not available message, but it gave a suggestion for something very close!

[chrome elk]

I'm looking to test my knowledge, anyone up for a challenge? It can be any kind.😅

[peak mulch]

https://tenor.com/view/high-five-fail-missed-high-five-high-five-fail-dumb-and-dumber-gif-7678007073368042247

Whew - what a journey to get into Discord. But I am here!

Is anyone frustrated with this non-descript error in Opal "Model generated a tool call but no tools were enabled in the request."

[odd widget]

I haven't hit that one yet but the system is still pretty limiting. I'm trying to use it for image generation it's slow + I can't figure out a way to make a loading screen while the user waits. Got any ideas?

[peak mulch]

it is frustrating - like it is random - I thought it was a ne project - but then a remix started doing exaclty the same thing. (assumed it was a API or key problem - I guess not)

[chrome elk]

What?

[quaint eagle]

Is anyone frustrated with this non-

I haven't hit that one yet but the

[chrome elk]

From what I understand, that error happens when the model tries to call a tool but none are enabled in the request. Basically, you need to enable the tools in the configuration so the model can actually use them.

[chrome elk]

From my many tests the first part isn’t possible, only after the application completes the flow. Of course, that’s just my observation.

I apologize for my English, I’m using Grok to translate for me.

[peak mulch]

@chrome elk yes - that is what I am not seeing in the interface - there does not appear to be a way to "set" or "configure" the tools. They are just offered in the interface as something to use.

[chrome elk]

So, in my observation, that part that shows up when debugging the application — I think that part can’t be changed because it’s not part of the app itself, but rather the application layer, you know? For example, I once made an application — I think I still have it — where the interface was just like ChatGPT’s, you know? But I could only get that information when it debugged all the way to the end.

[peak mulch]

yea - well in Opal - it is a bit frusating to get errors like this without any way for checking or configuring it... happy to click a button if I need to but there is no apparent setting

[chrome elk]

It’s because I think the idea behind it is more about low-code, drag-and-drop usage. There are many limitations you notice in it, especially when you want error log control. For example, you can’t implement an error log structure to know where the error came from or from which part it originated. There’s no such option, even if it’s done with AI. All the error logs it has are from the system itself.

[peak mulch]

agreed - I have never used n8n (seems a bit complex - more for developers who know a lot more than I do) -- but Opal seems like a nice balance for just prompting and flowing results from previous inputs. Brilliant in this respect / concept. I was kind of hoping for a bit more guiance on how to better deal with these kinds of erros and also better structure things. For Example: Look at this input - extract X, Y, Z for other operations later that I want to reference X, Y, or Z

[chrome elk]

Yes, at the moment it’s very limited.

[peak mulch]

I do like where they are going -- probably will not have outside Google usage (e.g. GPT, Grok, etc.) -- but for what it is .. .pretty slick

[chrome elk]

At this moment, I think it’s like this to see how far people will go.

[chrome elk]

Are there more groups that talk about opal?

[blissful bear]

Is there a way to embed output webapp from opal in another website ?

[frigid blade]

hi there! when will Opal be available in Romania?

[chrome elk]

I don’t know, friend.

[inland moss]

Went opal in Europe ?

[chrome elk]

West of Asia and north of Africa?

[tulip ravine]

yes

[chrome elk]

@sudden shadow

[odd widget]

For those who feel Opal is a bit limited right now (slow to use, doesn't start over, errors), you may want to try building the app in Google Studio (build tab). I tried it this morning and was able to accomplish the essential setup I was hoping for. You can see that at https://ai-hot-or-slop-497918404630.us-west1.run.app/

There are limitations. My main issue is if you want to use an outside service like Replicate / FAL for different models, don't do it. I've wasted hours trying to make it work and the app builder thinks it's my API key that's the fault (I have two other apps with keys that work without issue). Use Gemini services as much as possible but you'll run into quota limitations which will require you to get an API so Google will charge you directly. I understand Google has to get paid, but letting me know this ahead of time would have saved me hours.

There were other smaller issues but figured them out quickly since it's a small app. I think the general rule is that if you have over a couple thousand lines of code, VIBE coding on any system right now is very difficult (impossible if you don't have any coding skills).

[chrome elk]

Yeah, you need to manage the lines.

Are there many groups that talk about Opal? I really liked it and I’d like more.

[chrome elk]

Attention⚠️
⛔️Just a reminder⛔️ the idea and the code’s original construction came from @sudden shadow
I only took the idea and made an adaptation in Opal.

Credits to him.

If anyone needs help, you can DM me and I’ll assist you with her issue. Just so you know in advance, I don’t ask for money, passwords, emails, or anything like that. It’s just help.

[sudden shadow]

Thanks for the mention

[chrome elk]

@chrome elk The credit is yours ❤️

[elfin tulip]

it's top of mind for us, but right now no embeddings supported! What are some examples of where you'd want to embed Opals?

Can we remix our own after we've made it

[blissful bear]

folks will love to add small opals that enhance their websites. i guess it has to be added in an iframe which is ok but we would like to get rid of the opal header, hamburger menu etc.

[quaint eagle]

here's a silly one -- a tribute to Zombo.com
https://opal.withgoogle.com/?flow=drive:/1Y3DQ52hZGOMPY3UgZq51LpoUxVtUCRAz&shared&mode=app

[peak mulch]

I wonder how Gemma 3n would do as an optional model in Opal.... 🤔

[chrome elk]

I think someone deserves some food hahaha

[chrome elk]

That would be really cool tho.

new asmr video system

[dense totem]

Hello everyone! Similar to what we did last week, we wanted to share a quick update on some of the product improvements and new features we have worked on over the past week. Some updates are still work in progress, but we wanted to still give you a heads-up to keep you posted as we continue to move forward across our key focus areas.

🐛 Fewer errors with more resilient Opals: Opals are getting better at fixing themselves behind the scenes. Now we track faulty Large Language Model (LLM) calls, and re-try them under the hood. This means fewer interruptions and more seamless creative flows for you!

🔗 Easier path to share Opals with your team: You can now easily upload assets from your team’s Shared Drive and share your Opal with your team smoothly, with less permission headache.

⚡ Richer web outputs: We’ve enhanced Opal’s ability to generate richer, more interactive web content when using “Webpage with auto-layout” mode from the “Output” steps. Think of things like cool animated UI, interactive prompts, etc, all with slicker designs.

✨ No lag in the Remix button UI: We noticed an annoying lag on the ‘Remix’ button, which is now fixed. Building on others’ creations is now responsive and snappy, just as it should be!

⚙️ [Coming soon] Expanded Tools and support for Model Context Protocol (MCP). Many of you have been asking for expanded integrations with Google ecosystem (bringing Calendar/Gmail/etc into Opals), and connecting with 3P Tools via MCP. We are making great strides on this front. Stay tuned for what’s coming next to unlock some powerful new workflows!

[peak mulch]

1.29.0 (2025-08-15, 8:23 p.m.)?

[ashen badge]

Opal is really really similar to an app that I'm building...

[pine basin]

Will Opal be better than N8N?

[peak mulch]

At some point I believe it will be... right now I think n8n is more visual programming of workflows with detailed customizable interfaces betwen components/blocks. Opal does not appear to have that level of control... (yet)... and I am starting to wonder if it is needed the more I think about it. Context and data is moving in Opal vs specific hooks for operations. I think that is a good thing (edge cases will flip the switched on decisions between the two products)

I kind of view n8n as more deterministic and useful for specific structured data/operations (exclude the LLM factors that are fairly recent inclusions) -- there will always be needs for well defined workflows using structured data. But if Opal were to implement custom tools orchestrated by context -- that may be a game changer for deciding which tool to use for what purpose.

[chrome elk]

Yeah yeah, I agree.

[pine basin]

Great Analogy.

[chrome elk]

The arrival on Opal was such a wonderful and amazing thing that I was left without knowing what to do or having any ideas. I study prompt engineering every day, I’m a Java dev, and even so I have no idea what to do with Opal hahaha. The tool is so incredible that I just froze, I’ve been the whole day only wanting to have ideas and nothing… I’ve got the world in my hands and zero ideas hahaha.

[feral void]

Is it possible to create a tool that will scan my shopify store and choose products from a collection and generate a marketing email HTML code with the products and images and a brief description? I have come close but it always seems to use placeholder URLs and SRCs, which is weird because the description and price is correct.

[dawn hawk]

Why am I getting this error
{"code": "RESOURCE_EXHAUSTED", "message": "Quota exceeded for genmedia generation. Please try again later.", "details": "Quota exceeded for genmedia generation. Please try again later."} {"code": "RESOURCE_EXHAUSTED", "message": "Quota exceeded for genmedia generation. Please try again later.", "details": "Quota exceeded for genmedia generation. Please try again later."} {"code": "RESOURCE_EXHAUSTED", "message": "Quota exceeded for genmedia generation. Please try again later.", "details": "Quota exceeded for genmedia generation. Please try again later."} {"code": "RESOURCE_EXHAUSTED", "message": "Quota exceeded for genmedia generation. Please try again later.", "details": "Quota exceeded for genmedia generation. Please try again later."} {"code": "RESOURCE_EXHAUSTED", "message": "Quota exceeded for genmedia generation. Please try again later.", "details": "Quota exceeded for genmedia generation. Please try again later."}

This is the first time I'm using Opal and genrating any media

[elfin tulip]

Why am I getting this error

[quaint eagle]

Is it possible to create a tool that

[shy marsh]

Hi, when will Opal be available in Poland?

[gritty nest]

Hi, when will Opal be available in India?

[little shale]

VPN

[little shale]

VPN

[prime ibex]

hi when will opal be available in the Philippines?

[chrome elk]

@chrome elk

[chrome elk]

you are receiving this error because the media generation quota (genmedia) has been exhausted. You need to wait for the limit reset or increase the quota.

[chrome elk]

Today I helped two people, I’m happy; people used to laugh at me, and now they’re my friends.

I'm not buying anything

[chrome elk]

Well, analyzing the projects people make or try to make and all that, most of the projects where people get frustrated or can’t pull it off happen because they try to generate the video or the image inside OPAL itself. And I don’t recommend you do that. It’s really good for flow structures, for generating the final prompt. So the focus should be building your flow to generate the prompt that will then generate the video or the image.

Because it’s a really good flow tool, where you can create infinite flows, revisions, endless agents, whatever you want, you know? But when it comes to generating the video or the image inside it, I don’t recommend it, because many times what it generates isn’t even the prompt you created, you know? So if your goal is video and image generation, focus on prompt generation.

[peak mulch]

Very cool -- I almost thought it was a tilt-shift photo methodology... Would be cool if the people and vehicles were moving in these. as a prompt option...

[chrome elk]

Example. I have these two examples from the same project, where one of them involved video generation. It had two sections dedicated solely to video generation.

[peak mulch]

text is too fuzzy to comprehend. 🙁

[chrome elk]

So what did I do to fix it and improve the prompt that was already good? I just removed the video generation part. In this case, it had two sections only for generating video. And many times, the video wasn’t even being generated with the prompt that was made. So I took out all the video and music generation parts and put agents in their place. So instead of generating video and generating music, it’s agents reviewing, agents researching, and agents correcting.

[chrome elk]

That’s the intention haha

[chrome elk]

result

other examples

[peak mulch]

those are crazy cool. Well done!

[chrome elk]

That’s exactly it. So, if the main focus is video generation or image generation, focus the project on generating the prompt. Add as many agents as you want, both for reviewing and for researching. And remember, whenever you have a research agent, you must also have another one to review the research it’s doing.

[peak mulch]

agree - this is a solid approach. checks/balances.

[quaint eagle]

incredible quality.

[chrome elk]

Yes, yes, yes. It’s an approach I have myself. For example, if I have six research agents, I need to have six agents to review what was researched. Because just because the agent looks up something I asked it to research doesn’t mean it’s going to be correct.

[peak mulch]

HA - yes - this is the key. There is no "zero-shot" answer with a non-deterministic abacus at your fingertips.

[chrome elk]

Haha yeah bro

[chrome elk]

Another project completed! 🚀🔥
I finished the build based on an idea from @sudden shadow, who deserves all the credit for the inspiration 👏👏
Grateful for the reference, and on to the next ones!

reduced version

testing

result

[hollow linden]

what all can opel do?

[hollow linden]

.

[chrome elk]

Whatever I want, or imagine.

[chrome elk]

what will this be? 🧐

[peak mulch]

could go like 5x ways.

[chrome elk]

So, this was a play I started this morning, but I realized that, for now, insisting on it isn’t really worth it. I’ll get back to it at some point, but not right now.

[frosty thicket]

📣 We just deployed a new version of Opal 📣

It is expected that all currently signed in users will see an "Additional access required" prompt the next time they load. Please go ahead and sign in again, and make sure to select the new "See and download all your Google Drive files" permission when prompted.

The reason for this is that we are now requesting an additional permission. This permission substantially reduces friction relating to sharing. In particular, when using the "View Share Permissions" link to share your Opal with only specific users and groups, the people you share with with will no longer see the confusing "An Opal has been shared with you!" and "This Opal requires access to additional assets" dialogs. Files should also load a bit faster in general.

Thank you!

[chrome elk]

Wow, thank you so much! This is going to help a looooot 🎉

[quaint eagle]

https://tenor.com/view/tt-gif-22612306

Also in this release, we have two extra goodies:

• 💾 Download File option for your outputs, which saves the output as a file
• 📄 When saving outputs to Sheets/Docs/Slides, you get a nice clickable thumbnail of the saved document

[chrome elk]

Damn, is this already available?

[quaint eagle]

yup

[chrome elk]

Let’s run some tests.

[peak mulch]

@chrome elk thanks for the better images - they are easier to follow.

[chrome elk]

Haha, the next ones won’t be like that haha.

[oak hare]

is there no way for it to send the result to a custom api or something

[chrome elk]

So now you understand that this is the freedom we have?

Hshsh

[silent lynx]

anyone else getting this error? It is a youtube url processing remix, and I don't know what it means by caller????

[quaint eagle]

is there no way for it to send the

[pale axle]

we really need a central place to share gems/opal flows and anything else like that

[reef shadow]

is there anyone looking for dev?

[chrome elk]

💔

I'm only 12, agent

The other day I saw the idea of GoPro style videos, and I liked it.

[brittle osprey]

From Australia, just wondering how can we get access to Opal.

[chrome elk]

Vpn

[hoary star]

Or wait for official release

[royal night]

Hey everyone 👋
I’m building a tool for my agency that generates videos using Veo.
Here’s the workflow I want to implement:

User enters a prompt for video (+ optional dialogue + optional image).

As soon as they submit → credits get deducted immediately (no credit check).

The system builds a clean JSON payload for Veo and sends it via API.

Veo returns a video → we deliver it back to the user with a preview + link.

Dashboard updates credits accordingly.

Basically:
User Input → Deduct Credits → Send to Veo → Get Video → Deliver → Update Dashboard

I need help setting this up (API handling + credit deduction + error handling).
Anyone here who’s worked on similar flows and can guide? 🙏

[tired oracle]

Hi, do we already know when opal will be released or if there is an open beta or anything for EMEA?

[oak hare]

you cant link custom apis to opal atm

[chrome elk]

Dude, being real with you, swap this whole thing of generating videos for review agents and research agents. Like I said, it’s not worth locking yourself in, limiting everything to generation inside your setup, you know? So, focus on producing the prompt itself, you get it? Focus on producing the video prompt, the image prompt, you know? Because you’re limiting yourself way too much by sticking to generating video and image inside the setup.

[chrome elk]

My analysis of the platform, the project, and the framework is that it creates a lot of limitation, you know? Generating the video, generating the image inside the framework itself limits you too much, it basically takes away all the flexibility, all the freedom you have to create something bigger, you know? So from my point of view, bro, don’t focus on generating the video inside the framework.

[chrome elk]

Turn this into an agent, turn it into a research agent, a review agent. Set up agents that do the research on X topic and another agent that refines the result of that research. Add a review agent where it reviews that output. And if it doesn’t match the way the flow is happening, it asks to redo it again. So focus more on that, focus on redefining this idea through agents, because you have that possibility.

[chrome elk]

Look here, I’ve already shown this example, but I’ll show it again. This one is the same framework, these two examples are the same framework, the only difference in it is that I swapped the video generation part for agents that refine the final idea. And this is the result.

[royal night]

Got you bro, that makes a lot of sense. I’ll shift the focus to agents that do the research, refining, and reviewing instead of just locking it into video gen. That way it just builds the best prompts and I can run them anywhere.

btw can you please share that example you mentioned? Wanna see how you swapped the video part for refining agents.

[chrome elk]

here is the project with the generation parts

[chrome elk]

here with the agents, researching, reviewing, and refining

[chrome elk]

So buddy, your focus has to be on being able to structure your project to include agents — that’s the secret, aligning agents properly. And remember this: every time you add an agent whose job is to research something, you must add another one to review that research. Just because an agent found something doesn't mean it’s actually useful or exactly what was needed. So always pair research agents with review agents. If the result isn’t what was expected, the reviewer should request a redo. By keeping this flow, your final prompt will be more refined and detailed — because it’s not about how big the prompt is, but how refined it is.

[royal night]

Yeah that clicks bro 🔥 I see it now it’s not about making giant prompts, it’s about layering agents so they refine each other. Research + review as a pair makes the flow way more solid, and it keeps the final output clean and reliable. I’ll start structuring everything around that idea of alignment instead of just dumping tasks on one agent. Appreciate the gem🙇

[chrome elk]

Exactly, man you got my point. You need to expand your flow, you know? Not stay stuck in it. Generating the video or the image inside the project locks you down. You can’t be tied to the project. The project is just one step of the whole thing, you feel me?

[chrome elk]

Your main focus should be developing agents that can talk to each other where they build reasoning chains, review one another, process and improve the results. So in the end, you get a refined prompt, a refined output not just a giant prompt full of random info, but something sharp, clean, and well-polished.

[royal night]

Yeah bro, I feel you. I get it now the project itself is just one step, not the whole thing. The real focus should be on agents that can communicate, review, and refine each other’s work so the final output is sharp and polished. That makes way more sense than locking everything into video gen. Appreciate you breaking it down

[chrome elk]

If you need help, I'm available

[royal night]

Appreciate that bro I’ll definitely reach out if I get stuck. Good to know you’re around to guide me.

[chrome elk]

Exactly, that’s it I’m really glad you got it. It’s just one stage, you feel me? You can’t get stuck in the project just because it generates the video or image. That’s not the point. It’s a step a step where you refine your idea with agents, where you create agents that polish the final result. Don’t lock yourself in just because the project can generate media. That’s not the goal.

[royal night]

Yeah I get you bro, 100%. The media gen part is just a step, not the end goal. The real power is in the agents refining and polishing the idea so the final result is clean. Won’t lock myself into just the generation side 🔥

[chrome elk]

That’s exactly it, man. I’m really glad you got it. It’s basically like crafting a diamond ring. First, you need to cut the diamond, right? You refine it, shape it, polish it the whole process of turning that rough stone into something beautiful. And only then do you get the final result: the diamond ring.

[chrome elk]

new project

[chrome elk]

I blurred out some parts because of the nosy ones.

Haha

result

This DoubleFlux is based on a user from the channel. What it does is: you give your idea for an anime, a drawing, you know? Then it takes your description and redefines it, making your idea more refined. After that, it goes through two agents. One will do research about it, about the style, you know? It will gather that result, and the last agent will form the complete prompt. It will redefine everything, you know? And build a single refined prompt with all the information, but much more polished.

[chrome elk]

😮

[chrome elk]

entry into the flow :
Now she stands against a neon-lit rainy city backdrop, droplets catching ambient light around her silhouette. Her expression carries a soft melancholy, with lowered eyelids and a subtle downturn in her lips, reflecting the urban solitude around her.

[hollow wasp]

any user guide is there how to schedule to particular time

is it possible in opal

?

[feral void]

Are there any docs on how to get the most out of Opal?

[quaint eagle]

is it possible in opal

Are there any docs on how to get the

Thread

[elfin tulip]

Happy Friday from the Opal team! This week's product updates in time for your weekend Opaling:

🧠 A Smarter, More Reliable AI
The AI that powers our "Generate Text" feature just got a major upgrade. It's now better at thinking things through step-by-step, which helps it handle your requests more accurately. This means you'll see far fewer errors and get higher-quality results. We're also monitoring its performance so we can keep making it even better.

🚀 Easier Ways to Save and View Your Work
Download Your Files: You can now save outputs directly to your computer. Just look for the new "Download file" option.See Previews in Google Drive: When you save your work to Google Docs, Sheets, or Slides, you’ll now see a handy thumbnail preview of the file, making it easier to find what you're looking for.

🧹 Bug Fixes & Smoother Performance
We've been busy cleaning things up behind the scenes.

**Image & Video Bug Fixed: **We fixed an issue that sometimes prevented images and videos from being used correctly in your final output.
**Faster and More Responsive: **We've changed how the app processes your requests, allowing it to work on multiple steps at once. This makes the whole experience feel quicker and smoother.
**Minor Tweaks: **We've also updated the missing permissions pop-up and clarified our sharing disclaimer.

[zealous goblet]

Hi, can someone from the team or anyone (who has tried this) answer if we can build a Chrome Extension if so any video, guide or documentation, please.

[brave kestrel]

The Download your files feature is not working for me. The files it creates are unreadable. Am I missing something?

Interesting. I used a different Gmail log in and the "Download File" button is there!?

[mossy meteor]

I'm getting multiple errors trying to just extract information from a PDF - both when it is referencing in Google Drive and when it is uploaded as a file. Has anyone had this issue and found a way to have it work? The goal is I need the user to provide this input in the app as the PDFs will be different each time.

[mossy meteor]

Tried to take the concept down to the basics and pull from Google Drive and getting this

[chrome elk]

Hello friends, I'm back

any news?

[chrome elk]

I was away for 2 days and all this already happened, I already want to test it.

Any crazy idea for me to test?

[chrome elk]

Anyone got an idea? I’m dying to test it.

I'm coming back

nobody ?

[chrome elk]

some? 😭

[chrome elk]

https://tenor.com/view/rain-raining-winter-rainy-day-rainy-gif-17307471363181880117

[chrome elk]

testing, it turned out really good indeed

result

[chrome elk]

New 2D Sprite Creation Flow
Our system generates game assets in a Stardew Valley style, but with an astronaut theme 👨‍🚀✨.
The flow takes inputs like asset description, visual style, color palette, shape, lighting, level of detail, game genre, and sprite size.

🔹 It combines everything into a master prompt to generate:

Pixel art sprites in the chosen size (16x16, 32x32, 64x64).

Full spritesheets with animations (idle, walk, jump, attack, farming, fishing, etc.).

Environmental assets with dynamic animations (trees swaying, falling, water rippling).

🎮 This enables fast creation of an MVP game Stardew Valley in space with an astronaut farming, solving puzzles, fishing, and exploring planets.

[chrome elk]

hahaha

[chrome elk]

testing everything 💖

[chrome elk]

I wonder what that does, huh?
👀

[chrome elk]

Let me introduce you to my newest and most improved invention. I call it The Digital Drug.

[twilit charm]

when is opal for the non-us users? 🙁

[steep granite]

Hello everyone! For some reason, I keep getting the following error in the image to video tool - Video generation expects either a single text description or text plus a single image. Got 2 images. - but I am uploading only one image

[unborn raven]

When is opal scheduled for release in india

[heady dock]

i hadn't noticed this before, it might be new,
Opal now shows its thinking as it plans and builds your app

[lucid harness]

Soon.

[feral void]

I think I broke something

[quaint eagle]

Tried to take the concept down to the

The Download your files feature is not

I think I broke something

[chrome elk]

Vpn

I'm back with a new project

[chrome elk]

are you dreaming too big?

[twilit charm]

Yes I use it

[chrome elk]

still doesn't work?

[twilit charm]

Yes, it's work haha

[chrome elk]

need help?

[chrome elk]

👀

[quaint eagle]

Hello everyone! For some reason, I keep

[chrome elk]

For example, this same flow I showed that generated this video I realized I could make one more upgrade to it, which is what? Adding a secretary. So, in addition to having agents that refine the prompt, it will have secretaries to help them refine it. That way, I’ll get an even more refined result and be able to reproduce the quality in any AI.

[chrome elk]

with real-time feedback

[chrome elk]

credits to @final cairn prompt creator

[mental nebula]

Where can I find basic documenation for the Google Opal app? I am so lost on this workflow I'm trying to build and have no idea why it won't do the last step

[wintry bridge]

Can Opal be used to develop more complex UI app? Right now I don’t quite understand this product — it seems that Gemini Canvas is much more powerful. However, Gemini Canvas by default can only call 2.5 Flash, while Opal can call 2.5 Pro, is that correct?

I thought Opal was supposed to be something like “Gemini for app making,” but it doesn’t really seem to be that.

[wintry bridge]

Moreover, Opal’s AI generation and processing are extremely slow, often taking 10 to 20 minutes, and it is also prone to errors.

[silent lynx]

I think that is the path, at least I hope it is... I hope that we will be able to eventually make an Opal that has a user experience without exposing everything to the users and they will get the output (either on screen, or in Docs/Slides) but then I am just drinking coffee and dreaming

[chrome elk]

Man, I don’t know where it is either, I searched a lot but if you need any help, you know? You can talk to me, I can help you out.

[chrome elk]

Hey everyone, if I disappear or don’t reply to your questions, it’s because I’m focused on something right now.
I’ll be back soon and answer everything.
If it’s urgent, feel free to DM me! ❤️

[quaint eagle]

Where can I find basic documenation for

Moreover, Opal’s AI generation and

[quaint eagle]

I think that is the path, at least I

[stray acorn]

Heard about Opal, signed up and was told not available for my country.. well darn that feels rude after asking permissions for my google drives

Really would have loved to know that its only available somewhere

[chrome elk]

vpn

[quaint eagle]

Really would have loved to know that its

[mental nebula]

Thank you for the offer. Dimitri has also offered to help. I'll let him respond and then come back to you if that doesn't work. Thanks so much!

[inner parrot]

I made a tutorial video about JSON prompting that features an app built in opal, and got a message from someone who tried the link that they received this error message. Didn't know what to make of it as I had not come across any big problems myself. Is the error below a common one that's just part of a beta being a beta?

[marble mantle]

Hello guys, How i access the Opal from Bangladesh?

[inner parrot]

I think you need a VPN

[quaint eagle]

I made a tutorial video about JSON

[austere iron]

Question, if we create an app that allows for upload, is there a way to be sure that uploaded information is secure?

[elfin tulip]

Hey everyone!

Ealier today, Google announced our latest state-of-the-art image generation and editing model, Gemini 2.5 Flash Image model . **As a thank you to our early Opal users, we've made this newest model available to you all today! **✨ You don't have to opt-in; you'll see this model available for you in Opal immediately. Among other benefits, this new model should give you higher image quality, character consistency, and lower latency.

We're excited to see what you build with it! If you're so inclined to share the Opals you make with this new model, we'd love to see it here! Who knows, we may even be able to find a way to feature your Opal on a Google social channel 😉

[quaint eagle]

🖼️ Here's a simple demo of the new model: an image editor. Just give it a phrase and an image, and it will write the phrase over the image: https://opal.withgoogle.com/?flow=drive:/1C7wtjYznZU2GzxdksHnwlZPXu9ZJcNQi&shared&mode=app

[quaint eagle]

created with this demo

[chrome elk]

I got banned from the other server, right? I didn’t know it’s not in the rules about that, and maybe I’m just not meant to be there. So… goodbye.❤️

[quaint eagle]

here's another one -- Memeifier. Turns any picture you give it into a meme
https://opal.withgoogle.com/?flow=drive:/1IO3TN9YBWKNKkAWyKdIanlOl2Iw8bSzf&shared&mode=app

[peak mulch]

#Hilarious

Note - does not work with SVG (but I guess that is not a "picture") 🙂

@quaint eagle Pretty cool!

Very nice MEME.

Small glitch on letters.

the model does a good job relating the images.

[quaint eagle]

I wonder if there's prompt tuning that can be done to reduce letter glitches

[peak mulch]

Maybe just tell it to ensure all letters are properly displayed and words spelled correctly as part of your prompt? (force it to think)?

Sometimes it is not overlaying the image - but it may be deciding to NOT due that because there is not enough room. Maybe "out fill" if required <-- in the prompt? (e.g., extend the image boundary to include text?)

[vague coral]

Is anyone else having an issue with the theme generator menu not loading? The screen darkens on the editor but the menu isn't there.

[tepid moon]

https://opal.withgoogle.com/?flow=drive:/1sQKwgsZO1YM8xidto8r6ZQ3wgvfTqjFN&shared&mode=app

Hey my friend, I don’t know much, but I really love the concept of nanobots. So I added a few things to your flow based on that idea. Take a look and see if you like it!

[tepid moon]

That’s not quite the right approach, because otherwise you’ll overwrite part of the generation. You’ll end up replacing that part with this instruction.

[tepid moon]

Again, you’ll be overwriting an essential part just to include something so trivial, which could be structured in another way.

Let me guess. Your prompt is that one with a million lines, and you have no idea what each line does. And you have absolutely no understanding of what a structured block is.

Your prompt is massive, but you don’t even know the difference between the narration block and the sound block.

It’s just like people who say that stuffing the prompt with a huge JSON structure is going to improve the result or increase quality. If you have to rely on your prompt being overloaded with JSON to get quality, it means you're not actually doing anything.

Take your best prompt, the one that gave you the best result, in JSON format. I can rebuild it in a better way and get even more quality from it.

You don’t see it, but you’re trapped in the fear you create yourself by not having control of the situation. So you literally have no control over anything and stay stuck in it.🤷

[tepid moon]

I fixed that.

[tepid moon]

Here, it’s the same prompt in two versions, and the video generation is different. Can you understand what I mean when I say you’re stuck on this format issue, where you don’t have a solid structure?

[chrome elk]

I sent you a message on DM

My goal right now is to help you level up. So if you have any questions and tell me which language you speak, I’ll send you a voice message in your DMs explaining your problem or the solution to it in your native language.

At this first stage, it will be exclusive to Opal only. Later on, it will expand to other platforms.

anyone need help?

[chrome elk]

What do you think?

[chrome elk]

result

[chrome elk]

Hey everyone, just a quick heads-up for those who usually reach out to me for help or support. I was banned from the Google Gemini server, so right now I'm only active here. If you see anyone claiming to be me over there it's not me. Just wanted to make sure no one gets confused or tricked. Thanks!

[merry citrus]

maybe there's a way to appeal the decision?

[chrome elk]

I also want to know

If you find out let me know ok