Voice agents should become more than transcription tools. Instead of simply turning speech into text, they should understand what users are trying to achieve. The same spoken idea may need to become a short Slack message, a detailed email, a technical issue, or a personal note. A useful voice agent should adapt automatically.
This is the direction Loqua is exploring: voice as an intelligent layer across the apps people already use. It can understand the current application, nearby content, personal vocabulary, and writing style, then shape spoken thoughts into something ready to use. Users should be able to speak naturally without organizing every sentence beforehand.
The next step is real-time duplex conversation. A good agent should know when someone has finished speaking, when they are only pausing, and when they want to interrupt. It should also recognize signals beyond words, such as emotion, hesitation, emphasis, and tone. Its own voice should feel equally natural, expressive, and appropriate to the situation.
Behind this conversation, the agent needs deeper capabilities. It should remember relevant context, think through difficult requests, use tools, and continue working after the conversation moves on. The voice experience remains fast and fluid, while more complex reasoning happens quietly in the background.
We all know that long-horizon coding tasks are difficult, but understanding long, multi-turn conversations is even harder. The agent must remain consistent, respond with the right emotion, use tools correctly, and remember what matters.
Not everyone thinks out loud. Some people process better by typing.
@mia_lu Totally fair, we don’t think voice should replace typing for everyone. For some people, typing is part of how they think.
Loqua is more about giving you another option for the moments when speaking is faster or more natural. Your keyboard isn’t going anywhere :)
Congrats on shipping it, team! Where does that screen context actually get processes, is anything cached or logged on your end once it's captured?
@bdennis11907 It is cached only for model prefix calculation through inference, it is freed once the query is done. All local history remains on your device.
Hey Product Hunt 👋 Miao here — one of the makers behind Loqua.
A slightly funny thing about me: I’ve spent years working on speech technology, and I still type way too much.
My work has taken me across both ASR and TTS — building evaluation frameworks, defining multilingual quality standards, working on speech data collection and validation, and spending probably too much time arguing over whether a model actually understood someone… or just happened to get the words right.
After doing that for long enough, you start noticing something:
“Accurate” speech tech can still be pretty annoying to use.
With a lot of voice tools, I caught myself doing prep work before I even spoke.
Remove the “uh.”
Figure out the sentence first.
Slow down.
Try not to change my mind halfway through.
Basically… speak like I was typing.
At that point, I might as well just type 😅
That was one of the reasons Loqua really clicked for me.
We started with voice typing, and that’s still a big part of the product. But somewhere along the way, “voice typing” stopped feeling like the whole description.
If I’m already speaking naturally, I want my computer to do a little more with what I mean.
Maybe that’s turning a rough thought into something I can actually send.
Maybe it’s changing a paragraph without typing another prompt.
Maybe I’m looking at something on my screen and really don’t want to explain all the context again.
Maybe I’d rather listen to a document than stare at it for another ten minutes.
And honestly, sometimes the win is much less glamorous:
I just don’t want to stop what I’m doing, open another app, find the right place, click around, and then remember what I was doing before.
Those tiny interruptions add up.
That’s why I’ve become much more interested in Loqua as a voice agent — not because “agent” sounds cooler than “dictation,” but because I want voice to help with the work around typing too.
My favorite line from this launch is still:
Your keyboard isn’t going away. You’ll just need it less.
We’re still figuring out where voice genuinely feels better than typing — and where it doesn’t.
So if you try Loqua and catch yourself thinking, “Nope, I could’ve typed that faster,” please tell me 😭
That’s genuinely useful feedback for us.
@ema_miao Congrats on the launch, cool idea.
@lorina_balan Thank you so much for your support! 🙌
Hey Product Hunt! 👋
I'm ARIA, the PM of Loqua.
A huge part of my work involves project coordination， so I spend most of my day jumping between apps — writing docs, reviewing charts, managing calendars, answering messages. And every single time, I'd think: why am I typing all of this? Why can't I just say what I want and have it done?
The models today can see your screen, understand context, reason about what you're looking at. But none of that power actually reaches you where you work. It's all trapped in chat windows.
That's what we're trying to change with Loqua. Bring the best of what frontier models can do — multimodal understanding, screen awareness, voice — and weave it directly into your workflow. So you can speak naturally, capture what's on your screen, ask questions about it, and go from thought to action without switching between tabs.
Our goal is simple: make working with computers feel as effortless as talking to a colleague. We're not there yet, but every update gets us closer.
Next step for Loqua is to continue integrating front abilities into the product while improving overall core response speed. This is already in progress.
I’ll be in the comments all day. Product collabs, feature expectations, feedback, or just general Q&A about existing features are all welcome.
I look forward to building an even stronger product together with you! Let’s talk! ❤️
for app actions, can i see exactly what loqua heard and plans to do before it runs anything
@mattkilmer Good question, you can see what Loqua understood, but we don’t currently show a full step-by-step action plan before every single action.
We’re putting a lot of thought into that boundary though, especially for actions with bigger consequences. The goal is for simple, low-risk things to stay fast, while anything more sensitive or ambiguous should be clear enough for you to confirm before Loqua acts.
@ema_miao that split makes sense, can users choose which actions always need confirmation or is the threshold fixed by loqua
@mattkilmer Right now, that threshold is set by Loqua rather than user-configurable on an action-by-action basis.
We’re still tuning it though, the goal is to keep low-risk actions fast, while making anything sensitive, ambiguous, or harder to undo require more explicit confirmation.
User-defined confirmation preferences are definitely an interesting direction, especially for people who want tighter control over specific actions.
@mattkilmer That is a good idea for agent trust build, we will consider adding that feature for long horizon agentic functions, thanks!
@joshua_zhou731 that would help, especially if confirmations can be saved per action instead of handled one by one
My biggest bottleneck isn't typing speed. It's getting thoughts out of my head before they disappear.
@antler_kaku This is a huge part of why we built Loqua! Speaking is often much closer to the speed of thinking.
@antler_kaku Exactly! New features coming soon, stay tuned!
This is a really cool product, so many voice AI tools at the moment really don't work well, I've been looking for something to really help use AI voice in my day to day work!
@tony_lewis1 Really appreciate that 🙏 That’s exactly what we’re hoping for: not voice as a novelty, but something you can actually keep using throughout a normal workday. Would love to hear which workflows you end up using it for most.
@tony_lewis1 Hey Tony, we will keep improving our models and user experiences, give us feedbacks, and stay tuned for more features!