We built our own voice server
Yeonhong and Yeonha now send audio to Discord voice channels through Voicelink, a voice server we built ourselves.
The voice server we used before was built for music playback, so it was heavy and full of parts Yeonhong, which only reads TTS, didn’t need. We built a new one with only what TTS needs, and fixed some long-standing problems along the way.
- We fixed speech playing late.Sometimes the speaking indicator (green ring) lit up but the audio came a few seconds later. It now plays right away.
- The first sentence after rejoining is no longer cut off.We fixed the start of the first sentence being cut off right after someone rejoined a voice channel.
- The wait before each sentence is shorter.We streamlined how speech audio is fetched, cutting the small delay before each sentence.
- It’s lighter and more stable.The same server can now handle far more people, so drops during busy hours are less frequent.
Read Together and the evened-out silence between sentences in this update were only possible thanks to Voicelink.
Reading several people at once
When a chat moves fast, reading used to fall behind while the bot finished one person’s message before starting the next. Turn on Read Together and messages that several people send at almost the same time are read over each other. Messages from the same person are still read in order.
Turning on Read Together also turns on Split Left and Right. Overlapping voices are split between your left and right ears so it is easier to tell who is talking. The other side still plays quietly, so you won’t miss anything with only one earbud in.
Both settings are in /server settings. Read Together starts turned off.
Choose which messages get read
Instead of reading every message in the channel, each server can set its own conditions. Set them in /server settings → Reading Rules.
- Roles to read: only messages from people with the selected roles are read.You can pick up to 25. Pick a selected role again to remove it.
- Message prefix: only messages that start with the chosen prefix are read.The prefix can be up to 8 characters and is left out when reading.
- Maximum length: very long messages are not read.Set a limit between 10 and 200 characters. When a longer message comes in, the bot lets you know it can’t be read.
- Read Speaker Names: the bot reads who wrote the message first.When the speaker changes or 60 seconds have passed, it reads “name, message”. This one is in /server settings.
More natural reading
- Turn on Natural Voice and messages with several exclamation or question marks are read with more feeling.You can turn it on in /user settings. It works with voices marked 'Natural Voice' in the voice list.
- We evened out the silence before and after each sentence, which used to vary from voice to voice.The gap between sentences is now the same whichever voice you use, so reading sounds smoother.
- Deleted messages are not read.If a message is deleted while it is being read, reading stops right away. Edited messages are read as edited.
Custom Voice sharing is back
We have reopened Custom Voice sharing, which was unavailable for a while. Even if the creator deletes the voice, people who received it can keep using it. A confirmation window now appears before a voice is deleted.
Other changes
- We refreshed the bot’s wording and design.Messages are easier to read, icons replace emoji, and button colors are calmer. Settings panels now show results right inside the panel.
- We made /join smarter.If the bot is already in the same channel, only the person who called it is told. If it is in use in another channel, it asks first before moving.
- FixesWe fixed /skip sometimes skipping the next sentence too, some server settings not applying right after the bot restarted, and credits being deducted for sentences deleted before they were read.
The bot restarted briefly during the update, so it may have left your voice channel. If so, please call it back with /join.
Thank you for continuing to use Yeonhong. We will be back with more improvements.
