Hitoo - Speak all languages in your own voice, in real time
by•
Hitoo translates conversations in real time — not just words, but tone, intent, and cultural context — while preserving your own voice across over 50+ languages.
Replies
Best
Maker
📌
Hey Product Hunt 👋 Matteo here, founder of Hitoo.
A year ago, I was on a call with someone who didn't speak my language, using one of the "live translation" tools everyone knows. It worked — technically. But the person on the other side wasn't hearing me. Flat robotic voice, half a second behind, every bit of tone stripped out. We were exchanging information, not having a conversation.
That gap is the whole reason Hitoo exists.
Most tools do this: speech → text → translate → text → speech. Every hop throws away something — your voice, your intonation, the fact that you were joking. We rebuilt Hitoo as a direct speech-to-speech system that keeps your real voice and tone across languages, with enough context to follow an actual conversation instead of translating isolated sentences.
We first showed Hitoo here as a closed beta. Since then, we've rewritten the core: a new speech-to-speech architecture that preserves voice identity and cuts the delay down to something that finally feels like talking, not waiting.
It's still early and far from perfect, and I'd rather hear that from you than pretend otherwise. I'll be in the comments all day and would love to know:
— When did a language barrier last actually cost you something — a deal, a friendship, a moment?
— If you try the demo: whose voice did you test it with, and did it sound like them?
Thanks for checking it out 🙏
@matteo_pelosi - This tool is definitely something that people need today as they work across regions. For me personally it would be useful maybe 2 or 3 times a year. There is no pricing model details I could easily find on the site. Is it a monthly/annual subscription or will there be pay as you go options?
Best of luck with the launch.
Report
Maker
@codeandsea Thanks Brent, really appreciate you taking the time to try it out.
You’ve hit on something important: the “2–3 times a year” use case is exactly why pricing shouldn’t force a subscription on everyone. We’re planning pay-as-you-go as the default — you pay for the minutes you actually translate, no commitment — alongside subscription tiers for teams and heavy users who want predictable billing.
You’re also right that the pricing isn’t clear on the site yet, and that’s on us. We’ll get a proper pricing page up so people don’t have to ask. In the meantime, happy to walk you through the numbers directly if useful.
Thanks again for the thoughtful feedback and the kind words on the launch. 🙏
@matteo_pelosi - Thanks for the clarification on pricing, your model totally works for me.
Report
Tried it on a call with a Japanese client and was honestly shocked how well it picked up on tone, not just the words. The fact that it keeps sounding like me across languages makes it feel way less awkward than other translators I've used.
Report
Maker
@ramazan779050 This means a lot, thank you, Ramazan. Preserving tone and keeping your own voice across languages is the whole reason we built AURIS the way we did — most tools treat translation as just swapping words, and the result always feels a bit robotic and awkward. Hearing that it made a real client call feel natural is exactly the signal we're chasing.
Thanks again for trying it live on a real call. 🙏
Report
finally a translator that gets how my voice actually sounds when im tired or excited, the tone preservation feels way more human than other tools ive tried in 50+ langs
Report
Maker
@bayrakllar20010 This is exactly what we were hoping people would feel — thank you Mahmut. Tone is where most translators fall flat: they get the words right but flatten out everything that makes you sound like you, tired or excited or anything in between. That human quality across 50+ languages is the whole point of AURIS, so hearing it land means a lot. Thanks for putting it through its paces! 🙏
Report
The voice preservation across 50+ languages is genuinely impressive, most translation tools flatten you into a generic accent but this keeps you sounding like you.
Report
Maker
@veli342040 Thank you, Veli; that's the exact problem we set out to solve. Getting flattened into a generic accent is what makes most translation feel impersonal — you stop sounding like a person and start sounding like a tool. Keeping your identity intact across 50+ languages is the hard part, and it's the part we care about most. Really appreciate you noticing it. 🙏
Report
Tried it with a friend who speaks Japanese and the tone actually came through instead of sounding flat or robotic. The voice preservation bit is wild, it really does sound like me but in another language.
Report
Maker
@melahat45588 Love hearing this, thank you, Melahat! Japanese is one of the trickier languages to get tone right on, so the fact that it came through instead of sounding flat is exactly what we've been working toward. And yes — the voice preservation is the part that still feels a little magic to us too. Thanks so much for trying it out with your friend! 🙏
Report
the way it preserves your own voice instead of defaulting to some generic translator voice shows real attention to craft.
Report
Maker
@madanoglus4557 Thank you Sibel, "attention to craft" is a lovely way to put it. Defaulting to a generic voice is the easy path, and it's why so much translated audio feels impersonal — your voice carries who you are, and losing it in translation defeats the purpose. Really appreciate you picking up on that. 🙏
Report
the tone translation thing is actually wild, like it picked up sarcasm in a chat i was having with a french friend and kept it intact in english. kind of mind blowing for something running in real time.
Report
Maker
@nurgldurmalgvw Sarcasm is honestly one of the hardest things to carry across languages — it lives almost entirely in tone rather than in the words themselves, so most systems just drop it. Hearing that it survived a French → English conversation in real time is one of the best pieces of feedback we could get. Thank you Nurgül! 🙏
Report
Wow, that's really impressive! I'm curious how does the voice cloning work? Do I only need to register my voice once, or do I have to read a set of predefined sentences first? How do you synchronize and preserve the speaker's voice identity across different languages?
On enrollment: you only need to register your voice once — no reading a script of predefined sentences. A short sample is enough for the model to extract a speaker representation, and that's reused from then on.
On cross-language preservation: the key is that voice identity and linguistic content are handled as separate things inside the model. Your speaker representation is language-agnostic, so it can condition speech generation in any of the supported languages without being tied to the phonetics of the one you enrolled in. That's what stops you from sounding like a generic voice with an accent when you switch languages.
Happy to go deeper if you're interested in the technical side — always fun to talk to someone who asks the right questions. 🙏
Report
I know a startup that was accepted into YC that does this for the HealthCare industry. Do you specialize in a specific industry? or is this for the general public? I would imagine an industry-specific app would have a norower customer base but earn depth.
Report
Maker
@rachid_abadli Really good question, and one we've thought about a lot.
Our beachhead is CCaaS — contact centres, where multilingual support is a daily operational cost, and the ROI is immediate and measurable. From there we're expanding into regulated verticals like healthcare, legal and finance, where accuracy and confidentiality requirements are higher.
But the underlying bet is that this is infrastructure, not a vertical app. AURIS is our own end-to-end speech-to-speech model, so the same core serves a contact centre, a clinician, or a developer building on top of our API — the depth comes from the model quality, and the verticalization happens in the layer above it.
So: narrow entry point, general-purpose foundation. Happy to hear which industry you had in mind. 🙏
Replies
@matteo_pelosi - This tool is definitely something that people need today as they work across regions. For me personally it would be useful maybe 2 or 3 times a year. There is no pricing model details I could easily find on the site. Is it a monthly/annual subscription or will there be pay as you go options?
Best of luck with the launch.
@codeandsea Thanks Brent, really appreciate you taking the time to try it out.
You’ve hit on something important: the “2–3 times a year” use case is exactly why pricing shouldn’t force a subscription on everyone. We’re planning pay-as-you-go as the default — you pay for the minutes you actually translate, no commitment — alongside subscription tiers for teams and heavy users who want predictable billing.
You’re also right that the pricing isn’t clear on the site yet, and that’s on us. We’ll get a proper pricing page up so people don’t have to ask. In the meantime, happy to walk you through the numbers directly if useful.
Thanks again for the thoughtful feedback and the kind words on the launch. 🙏
@matteo_pelosi - Thanks for the clarification on pricing, your model totally works for me.
Tried it on a call with a Japanese client and was honestly shocked how well it picked up on tone, not just the words. The fact that it keeps sounding like me across languages makes it feel way less awkward than other translators I've used.
@ramazan779050 This means a lot, thank you, Ramazan. Preserving tone and keeping your own voice across languages is the whole reason we built AURIS the way we did — most tools treat translation as just swapping words, and the result always feels a bit robotic and awkward. Hearing that it made a real client call feel natural is exactly the signal we're chasing.
Thanks again for trying it live on a real call. 🙏
finally a translator that gets how my voice actually sounds when im tired or excited, the tone preservation feels way more human than other tools ive tried in 50+ langs
@bayrakllar20010 This is exactly what we were hoping people would feel — thank you Mahmut. Tone is where most translators fall flat: they get the words right but flatten out everything that makes you sound like you, tired or excited or anything in between. That human quality across 50+ languages is the whole point of AURIS, so hearing it land means a lot. Thanks for putting it through its paces! 🙏
The voice preservation across 50+ languages is genuinely impressive, most translation tools flatten you into a generic accent but this keeps you sounding like you.
@veli342040 Thank you, Veli; that's the exact problem we set out to solve. Getting flattened into a generic accent is what makes most translation feel impersonal — you stop sounding like a person and start sounding like a tool. Keeping your identity intact across 50+ languages is the hard part, and it's the part we care about most. Really appreciate you noticing it. 🙏
Tried it with a friend who speaks Japanese and the tone actually came through instead of sounding flat or robotic. The voice preservation bit is wild, it really does sound like me but in another language.
@melahat45588 Love hearing this, thank you, Melahat! Japanese is one of the trickier languages to get tone right on, so the fact that it came through instead of sounding flat is exactly what we've been working toward. And yes — the voice preservation is the part that still feels a little magic to us too. Thanks so much for trying it out with your friend! 🙏
the way it preserves your own voice instead of defaulting to some generic translator voice shows real attention to craft.
@madanoglus4557 Thank you Sibel, "attention to craft" is a lovely way to put it. Defaulting to a generic voice is the easy path, and it's why so much translated audio feels impersonal — your voice carries who you are, and losing it in translation defeats the purpose. Really appreciate you picking up on that. 🙏
the tone translation thing is actually wild, like it picked up sarcasm in a chat i was having with a french friend and kept it intact in english. kind of mind blowing for something running in real time.
@nurgldurmalgvw Sarcasm is honestly one of the hardest things to carry across languages — it lives almost entirely in tone rather than in the words themselves, so most systems just drop it. Hearing that it survived a French → English conversation in real time is one of the best pieces of feedback we could get. Thank you Nurgül! 🙏
Wow, that's really impressive! I'm curious how does the voice cloning work? Do I only need to register my voice once, or do I have to read a set of predefined sentences first? How do you synchronize and preserve the speaker's voice identity across different languages?
@avery_green Great questions, thank you Avery!
On enrollment: you only need to register your voice once — no reading a script of predefined sentences. A short sample is enough for the model to extract a speaker representation, and that's reused from then on.
On cross-language preservation: the key is that voice identity and linguistic content are handled as separate things inside the model. Your speaker representation is language-agnostic, so it can condition speech generation in any of the supported languages without being tied to the phonetics of the one you enrolled in. That's what stops you from sounding like a generic voice with an accent when you switch languages.
Happy to go deeper if you're interested in the technical side — always fun to talk to someone who asks the right questions. 🙏
I know a startup that was accepted into YC that does this for the HealthCare industry. Do you specialize in a specific industry? or is this for the general public? I would imagine an industry-specific app would have a norower customer base but earn depth.
@rachid_abadli Really good question, and one we've thought about a lot.
Our beachhead is CCaaS — contact centres, where multilingual support is a daily operational cost, and the ROI is immediate and measurable. From there we're expanding into regulated verticals like healthcare, legal and finance, where accuracy and confidentiality requirements are higher.
But the underlying bet is that this is infrastructure, not a vertical app. AURIS is our own end-to-end speech-to-speech model, so the same core serves a contact centre, a clinician, or a developer building on top of our API — the depth comes from the model quality, and the verticalization happens in the layer above it.
So: narrow entry point, general-purpose foundation. Happy to hear which industry you had in mind. 🙏