Viktor.comAn AI coworker that actually does the work
Promoted
XiangYinGe
Chinese dialect TTS you can try without signup
9 followers
Chinese dialect TTS you can try without signup
9 followers
Turn text into speech across 10+ Chinese dialects, including Cantonese, Sichuanese, Northeastern Mandarin, and Hokkien. Try 100 characters without signup, preview the result, and download the audio. Voice cloning, voice design, and APIs are also available.
The no-signup 100-character trial is a smart move, it lets you actually hear the dialect accuracy before committing anything. Cantonese in particular is notoriously tricky for TTS and it's impressive that it sounds natural instead of robotic.
@civitoglu6428 Thanks for trying the Cantonese voices, Nihat! The no-signup trial is meant to make it easy to judge pronunciation quality before paying. If you remember which voice or phrase you tested, I’d love to know — it would help me understand which combinations sound the most natural.
this looks super useful for anyone working with regional chinese audiences. one thing that would be really helpful is adding support for code-switching between mandarin and a dialect within the same audio clip, since a lot of real conversations mix them naturally
@kriyeino9 Thanks, Şükriye — that’s a great point. Mixed Mandarin and dialect speech is common in real conversations. XiangYinGe currently generates one selected dialect voice per clip, so I’d love to understand your use case better: would switching between sentences be enough, or do you need phrase-level switching within the same sentence?
A latency indicator during preview generation would be really helpful, especially since longer passages might take a moment. Showing roughly how much time is left before the audio finishes would make the trial feel less like a guessing game and let users plan their downloads more confidently.
@nurtenh3dn Thanks, Nurten — that makes sense. Generation time can vary by text length, dialect, and voice, so I’d prefer not to show a misleading countdown. A clearer progress state or elapsed-time indicator may be a better first step. Roughly how long did your generation take, and how many characters did you enter?
Could be great if you add a side-by-side player that lets me compare the same sentence across different dialects at once. Right now I have to generate each one separately to hear the differences, which makes it harder to pick the right voice for my project.
@okanmawx Thanks, Okan — side-by-side comparison fits the way people choose dialects and voices. Generating several versions on demand would also consume multiple credits, so I’m considering whether pre-generated samples or a 2–3 voice comparison mode would be more useful. Which dialects or voices were you comparing, and what kind of project are you working on?
Finally a tool that nails the Northeastern tone, which most TTS engines completely butcher. The Cantonese sample sounded a little flat compared to native speakers, but the no-signup preview is genuinely useful for quick tests.
@eyllyuguar3a3w Thanks, Eylül — I really appreciate the specific comparison. I’m glad the Northeastern voice worked well for you.
For the Cantonese sample, do you remember which voice and sentence you tried? When you say it sounded flat, was it mainly the intonation, rhythm, or expressiveness? That would help me reproduce the issue more accurately.