Would it be possible to consider adding a "real-time text display" feature?

Currently, the text isn’t displayed until after I finish speaking, so sometimes I’m not entirely sure what the system has recognized—especially when I’m speaking at length, I have to go back and review the entire transcript. The current recognition rate is actually quite high, and while I may not be familiar with the internal design, from a user’s perspective, being able to see the text generated in real time would provide greater peace of mind and allow me to spot recognition errors more quickly. This is just a personal observation based on my experience, for your reference.

建议能够支持语音输入的实时上屏，不然看不到进度总担心软件没有在工作，

对用户体验非常重要

[I'd like to see a new feature added] Support for streaming recognition

Enable real-time speech recognition with one click Now, during remote meetings with foreigners or while watching YouTube, speech can be recognized in real time and displayed in the client tool [Finally, the client tool can be put to good use!] Let AI correct errors while translating This effectively achieves simultaneous interpretation Please definitely implement this—it’s really needed 🫡

local user dictionary

We plan to develop the User Dictionary feature locally, and it will be very fast to run the User Dictionary feature locally.

本地模型可以接 Qwen3-ASR 吗？

Qwen3-ASR 听说效果非常不错，我蛮好奇的 https://huggingface.co/Qwen/Qwen3-ASR-1.7B 另外，建议不要整太多花哨的功能，就把语音识别做做好，甚至只把本地模型识别做做好，就很有特点了。

streaming output

Can we make this speech recognition a streaming output. The current experience with this one-time output doesn't feel very effective.

When I use Lightning, what I type is automatically saved to the clipboard. This feature isn't very good, and I hope it will be improved.

LightningSpeak works great—thanks. I’d like to report a minor issue, though: every time I use LightningSpeak, it saves what I said to the clipboard.But this feature isn’t very useful. For example, if I copy a piece of text first and then use Lightning Say, once I’m done with Lightning Say, the text I copied is replaced by what I just dictated, and the copied text is lost.This is very inconvenient to use, and since this feature isn’t really useful, it could be removed. I think many users probably copy some text first, then use Lightning Talk to speak the rest. But when they go to paste, the text they originally copied is gone. I find this very inconvenient.

local dictionary function

In the local user dictionary, can you do a quick import of the previous input method corpus

I would like to add multiple shortcut keys, each corresponding to a different prompt word.

Using multiple shortcut keys or combinations to trigger different prompts can significantly enhance usability in various scenarios.

鼠标模式建议。

可以自定义绑定按键，鼠标中键用的还是比较多的，我一般用来关闭浏览器标签栏，上下滑动等。如果能绑定在侧键就好了，现在鼠标很多都有上下两个侧键。

It was scanned by the company's IT and asked to be deleted.

Can some security policy be done so that the company doesn't think this is a risky application

Change the directory path for voice recordings

I hope you can add an option in the settings to change the directory for recording files. Alternatively, you could set an automatic deletion time—there really isn’t much point in keeping the recognized content once it’s processed. Or does the development team need to collect personal voice files for optimization? If that’s the case, I’m personally willing to share them, but please don’t hog all the space on my C drive!

希望能增加选项，让用户选择是否保留音频

个人不需要保留这个音频，更希望保留文本

Can it support mixed Chinese and English input?

Can you support Gemini?Lightning said to type it out :Can you support Jay Face Milk?

Can I sync multiple computers? I have quite a few computers—one at the office and one in my study.

I have a few more laptops. I'd like to sync my personal dictionary.

请允许用户定义models和recordings到其他硬盘

现在AppData/Roaming/Shandianshuo这个文件夹里含的内容占空间太大了，如果允许用户自定义把这个文件夹参与到D盘或者其他硬盘就好了。

录音不保存到硬盘

录音不保存到硬盘：所有操作都仅仅在内存中，不在硬盘中保留任何的文件（除非录音时间超过一定的阈值，如一分钟）

转写经常失败

我有很多次都是在直接说的时候，说发生网络错误，然后转写失败，然后就丢掉了语音。有的时候能转写过来，有的时候就转写不过来啊，这个很频繁呐，最近

希望增加多设备的配置同步功能

希望增加多设备的配置同步功能，目前词库、技能、模型配置等无法同步，尤其是词库无法同步比较麻烦

请求添加词库多端自动云同步功能

作者您好，辛苦了，没有什么bug，只是想提交一个希望做的功能。就是登录账号后，能自动同步各端的词库（词典），尤其是我这种mac Windows ios Android都有的人，每个软件单独同步词句太麻烦了！这个不切实际的，也不方便，希望能加个这个功能。以后手机端在说话过程中觉得有问题，我随手就录入软件词库了，希望电脑上或者其他端能同步这段录入，这样就会方便舒服很多。

​【Bug】自定义语音识别服务商接口路径被写死拼接 /chat/completions，导致 OpenAI 兼容 ASR 完全无法连接

更新软件导致配置丢失

你们这是什么草台班子，更新到0.7.5，配置丢失，所有模型的key、配置都要重新弄。已经发生过3次。就这么个基本逻辑硬是弄不好？

会有移动端的开发计划吗

Is corporate use permitted?

Hello, I’d like to install Flash on my work computer. Is this permitted? I haven’t been able to find a license or any related instructions, so I’m asking here.

Looking to reduce memory usage now that the model is memory resident (~1.9GB macOS)

Memory usage is still a bit large and takes up a bit too much performance on a normal computer, see how it can be tuned down. Is it possible to support remote invocation (e.g. allow us to deploy the model on our own server and get the text via callback). Or provide more lightweight models that are optional.

用户词典不起作用

我平时会输入一些专业的学术名词，但是我发现在记忆-用户词典里面输入了很多词汇，但是并不起到效果。

Can I simply pay a fee to avoid configuring the voice model myself?

Although I can configure the model myself, it’s still a hassle, and I have to set up a large language model service. Long-term use of a large language model platform inevitably incurs fees. Could you consider supporting a subscription model similar to Typeless, where users don’t have to configure or pay for the large language model themselves, but can simply subscribe and pay directly within Shandianshuo?

能否支持Gemini

可以支援繁體中文輸出嗎

0.6.0 Mac版本有明显吃首字问题

升级了新版本，多次发现有吃首字的问题。即按下快捷键后说话，发现录音一开始的两个字没有被识别到。怀疑是改了快捷按键响应后，导致录音启动延迟增加了。

有没有发现，开头几句总是识别不清楚

体验了1天，感觉每次识别中，开头几句都识别不清楚，后面的准确率还是比较高的 不知道有没有工程方法来优化下

Traditional Chinese Support

I hope to support traditional Chinese output. All Chinese characters will be output in traditional Chinese for the convenience of users in Hong Kong, Macao and Taiwan.