Lispr

• 1. overview (624 chars)

  Lispr is a free desktop application for voice dictation and real-time translation. Its main feature is "press and hold a key to speak, and the text will appear at the cursor when you release it." It was built by the Ukrainian software company Codebridge. It logged on to Product Hunt on July 9, 2026 and won the "fifth place of the day" product (about 290 votes). The biggest differentiation of the product is that "press and hold the key to dictate, and add another key to translate while speaking." It is completely free, no registration is required, and the installation package is only about 4 MB. For people who type in more than a dozen applications every day and have multi-language communication needs, it is a low-friction voice input solution.

• 2. Product background (893 chars)

  Codebridge is a Ukrainian software development company that engages in profitable software consulting outsourcing business. Lispr’s co-founder and CEO is Konstantin Karpushin, and co-founder and CTO is Myroslav Budzanivskyi. Konstantin said in his Product Hunt bio that the first user of Lispr was himself - his workday was filled with Claude Code sessions, customer emails, Teams discussions and document reviews. When he switched to using voice instead of keyboard, his output immediately doubled several times. The team itself is a development company, so we just did it ourselves.   The multi-language function also has a personal touch: the team is Ukrainian, speaks Ukrainian internally and English to customers, and many people live in a third language environment, so the translation is assigned a separate button. There is currently no external financing record in public information. Codebridge relies on consulting business to support Lispr, and claims that its pay-per-call architecture is extremely low-cost and can be free for a long time.

• 3. Product functions and usage experience (1877 chars)

  The core interaction is minimalist: In any application with a cursor, press and hold the right button (right Option key on Mac, Ctrl on Windows) to speak, and when you let go, the transcribed text will be inserted at the cursor position; if you hold down the Control key (Mac) while speaking, the translated text will be dropped when you let go, not the original text. It can work in Slack, email, Notion, Figma, terminal, browser input box and other places where there is a cursor, and supports switching languages ​​in the middle of a sentence - you think in Ukrainian, and the dropped message can be in English. It supports dictation in about 99 languages, translation in about 32 languages, and automatically detects languages.   Speed ​​is its main selling point. The officially measured median delay from "key release to text on screen" measured on real traffic is 346 milliseconds, and the dictation itself is about 200-300 milliseconds. The implementation method is to start streaming the audio to the cloud Whisper large-v3-turbo model (via Cloudflare edge proxy to Groq) before finishing speaking, and route it according to the nearest area where the user is located to ensure that the delay in different regions is stable. The entire application installation package is only about 3.67 MB, does not require downloading any model files, and does not require a local GPU. It can run on macOS 11 and above (including Intel and Apple Silicon) and Windows.   It also comes with a custom thesaurus function: you can teach it to recognize proper nouns such as customer names, product names, code identifiers, etc. The dictation and translation engines will respect this thesaurus and avoid "correcting" brand names into ordinary dictionary words. The thesaurus is saved in a local file on the machine. It will automatically review recent dictation every day and correct misheard professional words with correct spellings. It can also be added or deleted manually. In addition, it will be restored by the clipboard content after inserting text, and will not destroy what you copied before; the application itself has passed Apple notarization, and the Windows version is SmartScreen certified and code signed by "Codebridge Technology, Inc."

• 4. Pricing and business model (637 chars)

  Lispr is currently completely free, with no accounts, no subscriptions, and no paid plans. Konstantin clearly stated on the release page that "the free files will always be retained." Codebridge is a profitable consulting company. Lispr's architecture pays per call. Infrastructure costs follow usage and are not resident on the GPU, so the cost of maintaining free is very low. He promised that if a paid plan is added in the future, it will only be for heavy or team-level usage and will never affect daily dictation.   It should be noted that some third-party navigation sites have marked prices such as "one-week trial" or "starting at $29/month", which is inconsistent with the official statement. The official website should prevail - currently it is free.

• 5. User reviews (1372 chars)

  Lispr on Product Hunt received 3 formal reviews, with an average score of 4.67/5. Positive feedback focuses on three points: first, it is fast, with dictation and translation almost instantaneous; second, it can be used in any application, including terminals and Figma; third, it is completely free, with no account or subscription. A marketing team member said that she uses it to draft Slack, emails, and documents almost every day. She said that it does not change her behavior in different applications, but it just makes it faster to say the entire paragraph with context. Some users mentioned that when giving instructions to AI tools such as Claude and Cursor using voice, they can give two minutes of context and constraints in one breath, and the quality of the answers they get is significantly higher than the three-sentence streamlined prompts when typing - this was the original motivation of the founder to create this tool.   In terms of negatives and concerns, some people in the community discussion also mentioned several real pain points: translation and dictation are dependent on the Internet, and there is no offline mode; it is mainly based on macOS experience (although it is officially supported by Windows, but most early users are Mac users); it does not support advanced editing such as punctuation management and voice commands; the recognition accuracy is affected by the microphone and background noise; the vocabulary currently only exists on the local machine, and will be lost if you change to a new Mac or reinstall. The export/import function team has scheduled it but it has not yet been launched online.

• 6. Industry evaluation (745 chars)

  In the Product Hunt daily list on July 10, 2026, Lispr appeared together with new products such as Toyo and Auriko, and was rated by the Chinese technology aggregation site as an eye-catching tool that is "free, does not store audio, supports 99 languages, and has an average delay of 346ms." Jeremy Caplan of Wonder Tools lists it as a free alternative to Wispr Flow for what he calls "bionic dictation" - speaking your thoughts out loud before the urge to edit comes to you.   It is generally compared with Wispr Flow ($15/month), Apple native dictation, and Windows Voice Typing on third-party evaluation sites (such as ToolRadar, AIPure, and aitolly). The conclusions are consistent: Lispr's free, account-free, and built-in translation are its main advantages over subscription-based and system-built-in solutions. Its shortcomings are offline capabilities and advanced editing.

• 7. Controversies and Risks (945 chars)

  Privacy was the most scrutinized part of the discussion. The official caliber is: the microphone is completely turned off before pressing the button; the audio is streamed through an encrypted connection and discarded after transcribing. Lispr's own server does not store or record any transcribed content. But there is one detail worth noting: for the purpose of abuse review, the inference service provider (i.e., the cloud vendor that actually runs Whisper) will retain the audio for up to 30 days before deleting it. This is a slight tension with its "never store" statement, and it is a boundary that users need to be aware of.   In addition, although the thesaurus only stores local files and has no accounts, the balance between "the application remembers your lingo" and "no persistent identity" relies on local files; it starts as soon as the device is changed, and there is currently no migration method. Another point: it is just speech-to-text conversion without any rewriting or polishing. Punctuation and alignment completely depend on what you say. Long formal manuscripts still need to be organized by yourself.

• 8. Applicable groups and usage suggestions (864 chars)

  Suitable for: People who type a lot every day, developers and writers, customer service and multi-lingual communicators, people who use voice to give complex instructions to AI tools, people who need to drop foreign language messages directly into the chat window, and people who want to use "speaking" instead of "typing" to organize their thoughts. Not suitable for: Scenarios that require complete offline operation; environments that have extremely high requirements for data localization processing and inference service providers will not accept even a short stay; people who want to get a complete manuscript with punctuation and formatting at once, but do not want to polish it manually.   As an alternative, if you are looking for a free native experience, choose Lispr; if you are used to the subscription system and want more complete editing and punctuation management, you can look to Wispr Flow; if you only use it locally and don’t mind not having translation, you can also use macOS/Windows’ built-in dictation.

• 9. Summarize (383 chars)

  Lispr makes a complex task simple again: free, no account required, about 4 MB, press and hold the key to speak, the text (or translation) will fall to the cursor, and the delay is reduced to less than half a second. It may not be the most versatile dictation tool, but it is the one with the lowest threshold and the most convenient one. It is especially suitable for people who use voice to give instructions to AI and seamlessly switch between multiple languages.

No in-depth report yet

• 

  ChainChaserKristensen

  —

  用了几天 Lispr，最大的感受就是「轻」—— 安装包才 4MB，按住 Option 键说话、松手文字就出现在光标处，比我想象中自然得多。之前用 Wispr Flow 总觉得 Electron 那套太重了，启动慢半拍，换到这个完全没那感觉。

---

• 

  Dorothy538

  —

  翻译功能确实实用，跟国外同事在 Slack 上沟通，我说中文、松手出来的是英文，不用再切到 DeepL 来回粘贴。延迟大概半秒多吧，可以接受。要是离线也能用就好了，偶尔在飞机上还是想用。

---

• 

  MMurray520

  —

  说实话，比 Apple 内置的语音输入强太多了。系统自带的那个你说话得一个字一个字蹦，Lispr 你可以一口气说完一整段，识别率还高。最骚的是边说边翻译，按住 Ctrl+Shift 直接出英文，不用 App 间切来切去。

---

• 

  Hnflo

  —

  装了三天，已经回不去纯打字了。给 Claude 写 Prompt 以前要打半天，现在按住说话就完事，上下文给的足，回复质量明显更高。感觉这东西对开发者来说是个隐藏的效率神器。

---

• 

  re630tw

  —

  免费的工具能做到这个程度真的良心。之前试过 Wispr Flow，免费版一周就 2000 词，用两天就见底了。Lispr 目前完全免费，而且没账号注册，装好就能用。开发者说早期访问期间不收费，希望一直保持。

---

• 

  飞鸟922

  —

  346ms 延迟基本感觉不到，比我想的还快。松开按键文字就已经在屏幕上了，比某些商业软件还利索。说真的，不敢相信这是免费的东西。

---

• 

  CWhite_Max5

  —

  要能支持中文方言就好了，比如粤语。有时候跟家里人说粤语，Lispr 就听不懂了，识别直接崩。不过这个要求可能有点高，毕竟 Whisper 对方言的支持本身就有限。

---

• 

  m6yhl5ymhl

  —

  好奇有人在 Figma 里用过吗？我试了下在编辑设计稿的时候语音输入文案，居然可以直接在文本图层里打字，不用换窗口，体验挺丝滑的。不过切换中英文输入法的快捷键偶尔会和 Lispr 冲突，改了下触发键就好了。

---

• 

  lazyduck178

  —

  有个问题就是翻译的准确率偶尔会翻车，尤其是遇到行业术语或者比较口语化的表达。比如我跟供应商聊一些产品规格参数，翻译出来的结果有时候要再手改一下。日常沟通没问题，涉及专业内容还是得自己过一遍。

---

• 

  beautifulbutterfly237

  —

  吐槽两点：第一，没有离线模式，没网就废了。第二，不能自定义标点符号风格，有时候我想让它自动加逗号、句号，但它只会按默认规则来。希望后续版本能加个标点模式选项。

---

• 

  Ralph.ReyesIII

  —

  跟竞争对手比了一下，Lispr 最大优势就是免费+轻量。Wispr Flow 要 15 刀一个月，Superwhisper 虽然本地运行但是配置复杂。Lispr 不用注册不用下载模型，4MB 搞定一切，对大多数人的日常使用来说完全够用。

---

• 

  吴莉

  —

  翻译延迟没有官方说的那么低，我实测大概 1 到 1.5 秒。纯听写确实很快，300ms 左右，但加上翻译就明显有停顿感，说一句话等一秒，节奏被打断。不过考虑到完全免费，这个水平完全可以接受。

---

• 

  KEgon

  —

  意外发现写东西的时候用 Lispr 反而更流畅。平时打字思维会跟不上手速，现在直接说出来再改，思路更连贯。写周报、写邮件现在都是先语音输入再微调，效率翻倍。

---

• 

  qt588kc

  —

  有意思的是这个 app 的隐私策略做得挺到位的。音频加密传输，转录完立即丢弃，不存服务器，不用来训练模型，而且不需要创建账号也就没有个人信息泄露的风险。在语音工具里算是最干净的那一波了。

---

• 

  Maria.Henderson_X

  —

  测试了下中英混说的场景，我说「let's ship this feature 但是还需要加一些测试」，转录出来基本完整保留了两边，中英文都能正确识别。Apple 自带那个根本做不到这种 code-switching 识别，这点确实强。

---

• 

  SCollins_9903

  —

  做跨境电商的表示这东西确实能提高效率。以前给巴西供应商写葡萄牙语邮件，得先写英文再贴到 Google Translate 翻译完再复制回来，现在按住 Ctrl+Shift 说话直接出葡萄牙语，一步到位。不过翻译质量偶尔需要手动修一下。

---

• 

  Carol.Miller47

  —

  Windows 版用户路过，安装包大概 8MB，比 Mac 版大一点但还是很轻量。按住右 Ctrl 说话就行，操作逻辑一致。唯一问题是 WhatsApp Web 里偶尔会插错位置，光标跑偏，目前还是靠 Slack 渠道沟通为主。

---

• 

  MargaretTorres1687

  —

  下载了两个小时就离不开了。按住说话松手打字，这种交互真的太 intuitve 了。唯一后悔的是没早点发现这个工具，之前一直忍着用系统自带的语音输入，差距真不是一点半点。

---

• 

  Margaret.GarciaSr

  —

  开发者是乌克兰的团队，这点挺让人有好感的。看 Product Hunt 上的介绍，他们自己就是多语言团队，内部用乌克兰语、对外用英语、还有些人在第三国生活用第三种语言，所以翻译功能是他们真正需要的产品，不是那种随便加的功能。

---

• 

  MAree

  —

  我可以自定义词典，把一些产品名和代码里的 identifier 加进去，转录准确率提升不少。之前在终端里说 git 命令相关的内容，Lispr 老是识别错，加了自定义词汇后就稳了。

---

• 

  Douglas_Cox_7

  —

  趁 Early Access 赶紧用起来，不知道以后会不会收费。看了他们的说法是架构按调用付费，免费版应该能一直维持下去，但如果未来加付费套餐我也能理解，毕竟服务器成本在那。

---

• 

  MsAmeliaPicard_x

  —

  UI 做到极致了——根本没有 UI。打开就是个菜单栏图标，按住键说话完事。没有弹窗、没有订阅提示、没有多余的设置页面，这才是工具该有的样子。现在太多软件做成全功能的庞然大物，Lispr 这种专一做事的产品反而难得。

---

• 

  KAtur

  —

  给 Cursor 下指令特别爽，以前打字会偷懒只写三句话，现在抱着说两分钟上下文，Claude 回来的东西明显更对路。这玩意儿最值钱的就是让我对 AI 不吝啬话。

---

• 

  常怡涛

  —

  我是个独立开发者，基本泡在 AI 编码会话里。「打字让我舍不得把上下文告诉 AI」这一点太准了——我之前为了省按键，给 Cursor 的提示词都缩成一两句。换成 Lispr 之后，我能一口气把试了什么、哪里报错、哪些约束重要全说出来，Agent 不再瞎猜。关于代码术语它处理得也不错，专有名词用「词库」教一遍就记住了，连跨语言都说母语它也能打出正确的拉丁拼写。

---

• 

  DEpat

  —

  免费的语音输入，真香。

---

• 

  悠然568

  —

  在 Windows 上试了电商客服场景，整体还行，零配置、本机存配置方便团队分发。但翻译延迟实测 1.2 到 1.8 秒，bilingual 说话时有点卡顿；WhatsApp Web 里偶发插到错误光标位置。给供应商发短消息够用，正式文案还是得人审。

---

• 

  2ertm2vez

  —

  唯一不爽的是必须联网，地铁里基本废了。

---

• 

  GEpet

  —

  装完 4 MB，秒下。

---

• 

  Scott_Henderson3697

  —

  试用了一天，在 Slack 和 Notion 里直接说话出字，比切输入法快多了。

---

• 

  云朵_2

  —

  隐私这块问清楚才敢用：麦克风按键前是关的，音频加密上传转完就丢，官方不存转写内容。但有意思的是推理服务商（实际跑模型的云厂商）会留音频最多 30 天做滥用审查，这点官网没大张旗鼓说，介意的要注意。

No reviews yet