Lispr
| Developer(s) | Codebridge Technology, Inc. |
|---|---|
| Initial release | May 17, 2026 |
| Engine | |
| Operating system | macOS, Windows |
| Type | Speech recognition, machine translation |
| Website | lispr |
Search Lispr on Amazon.
Lispr is a voice dictation and translation application for macOS and Windows, developed by the Ukrainian IT company Codebridge.[1][2] It converts speech into text, inserts the text at the cursor in any application, and can translate dictated speech into 32 languages.[1]
Development
Codebridge, a software development and outsourcing company founded in 2021, built Lispr as one of its own products. The company used a multi-agent AI system for coding, testing and localization during development.[1] Lispr became publicly available in early access in May 2026.[3] The macOS client is a native application written in Swift, and the Windows client is built on .NET and Windows Presentation Foundation.[1]
Features
Dictation
Lispr works on a push-to-talk basis. The user holds a trigger key, speaks and releases the key, and the recognized text is inserted at the cursor in the active application.[1][2] It works in any text field, including applications such as Slack, Figma, terminals, text editors and integrated development environments, and restores the previous contents of the clipboard after inserting the text.[1] The trigger key can be changed to another modifier key or to a function key.[4]
A hands-free mode starts recording with a double tap of the trigger key and stops it with a single tap.[3][4] A single recording can last up to two hours and can stop automatically after a period of silence.[3]
Speech recognition runs in the cloud on a hosted version of the Whisper large-v3-turbo model, so the application needs an internet connection.[1][4] The spoken language is detected automatically from about 99 languages.[5] The developer reports a delay of about 1/3 of a second between releasing the key and the text appearing.[1]
Translation
Translation was added in May 2026.[3] When the user holds a second key together with the trigger key, Lispr inserts a translation of the speech in place of the transcript, in one of 32 target languages, including Ukrainian, English, German, Spanish, Japanese and Arabic.[1] Target languages are assigned to two translation slots, each with its own key.[4] The developer reports a translation delay of up to about 0.7 seconds.[1]
Vocabulary and formatting
Users can build a custom vocabulary of names, product names, specialist terms and code identifiers. The speech model receives the list as a hint before each dictation, and the list is applied to both recognition and translation.[6][1] One vocabulary covers every dictation language and accepts entries in Cyrillic, Hebrew and CJK scripts.[6] An automatic mode, enabled by default, scans recent dictations at most once every 24 hours and adds new terms to the list. To do this, it sends excerpts of recent dictations, capped at 500 characters each, to the company's server for analysis.[6][3] On macOS, the vocabulary is stored as a local file and is not tied to an account.[6]
A formatting setting offers three modes. Off leaves the transcript unchanged, Light adds punctuation, paragraphs and lists, and Concise also removes filler words and repetition.[4][3]
History and privacy
A history view lists past dictations and lets the user retry a dictation that failed.[3] Dictation does not require an account. An optional account, used for paid plans, signs in with a one-time email code.[4][7] According to the company's privacy policy, audio and transcripts are not stored on its servers by default, dictation history stays on the user's device, and audio and transcripts are not used to train models unless the user turns on an optional sharing setting. Audio is processed by a third-party transcription provider, whose own privacy policy applies.[7]
Platforms
The macOS version requires macOS 11 or later, runs on both Apple silicon and Intel processors, appears in the menu bar and the Dock, and is notarized by Apple.[8][5] The Windows version requires 64-bit Windows 10 or later, runs from the system tray, and is code-signed through Microsoft's Azure Trusted Signing service with Codebridge Technology, Inc. as the verified publisher.[8][1]
Reception
In July 2026, Jeremy Caplan recommended Lispr in his newsletter Wonder Tools as a new alternative to the dictation app Wispr Flow. He reported testing it by dictating messages and translating an email reply into Portuguese, which he said worked well.[9] Later that month, the Ukrainian IT community site DOU recommended Lispr as an AI translator in its weekly news programme DOU News.[10]
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 1.10 1.11 Kuzmenko, Oleksandr (14 September 2026). "Ukrainian IT company Codebridge has developed a voice application Lispr with instant translation into 32 languages". dev.ua. Retrieved 30 October 2026.
- ↑ 2.0 2.1 Katerynchak, Yevheniia (17 September 2026). "Підтримує понад 30 мов: українці створили застосунок з миттєвим перекладом" [Supports over 30 languages: Ukrainians created an app with instant translation]. tykyiv.com (in українська). Retrieved 30 October 2026.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 "Changelog". Lispr. Codebridge Technology, Inc. Retrieved 30 October 2026.
- ↑ 4.0 4.1 4.2 4.3 4.4 4.5 "Help". Lispr. Codebridge Technology, Inc. Retrieved 30 October 2026.
- ↑ 5.0 5.1 "Lispr". Codebridge Technology, Inc. Retrieved 30 October 2026.
- ↑ 6.0 6.1 6.2 6.3 "Teach Lispr your words". Lispr. Codebridge Technology, Inc. 30 May 2026. Retrieved 30 October 2026.
- ↑ 7.0 7.1 "Privacy Policy". Lispr. Codebridge Technology, Inc. 23 August 2026. Retrieved 30 October 2026.
- ↑ 8.0 8.1 "Download Lispr". Lispr. Codebridge Technology, Inc. Retrieved 30 October 2026.
- ↑ Caplan, Jeremy (4 July 2026). "A Guide to Thinking With AI". Wonder Tools. Retrieved 30 October 2026.
- ↑ Kovalevskyi, Yevhen (13 July 2026). "DOU News #258". DOU (in українська). Retrieved 30 October 2026.
This article "Lispr" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Lispr. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
