Hey, I would be a lot faster if I could just speak to the a composer/chat and it translates the speech to words.
I guess youâre referring to STT (speech-to-text) instead? Being able to speak your prompt instead of typing?
â Definitely needed! Though, there are external solutions and applications to this, or you could even build your own in the meanwhile - best of those are platform-specific however (Apple)..
1 Like
Well really, TTS (model-to-user) and STT (user-to-model): I would like to have a conversation with the agent. In a sense, I think a key metric for cursors success is the split between the agent window and the code window. The smaller the code window per unit of time spent, the better cursor works.
For example, I would like to talk to the AI while I am traveling. Plan out the app, check implementation variations, run tests etc. Then when I come back to the laptop, I have a summary of what was discussed: a nicely organized dashboard. So then I can simply hammer out the various approaches on the laptop.
The TTS-STT is needed because a) I canât type the complex instructions into my mobile when afk and b) I program now in language so no need for the precision - I need speed and language is faster than typing (and more convenient).
Just think of it like the OpenAI real time API with a great code interface and dashboard.
I would like to implement a screen reader (TTS) if I can have an API (or even just applescript) access for that. Any interest?
Iâve been using SuperWhisper flawlessly for this and can highly recommend it.
Feature request for product/service
Chat
Describe the request
Hi team,
Iâd like to request a speech-to-text feature in the chat window. The idea is to allow users to speak their prompts instead of typing, and have Cursor automatically convert the speech into text.
This would be extremely helpful in situations where:
A user has an injury or temporary limitation (for example, my left hand is currently fractured, and typing long prompts is slow and painful).
Users want to quickly jot down thoughts or ideas verbally without breaking their flow.
Accessibility needs, making Cursor more inclusive for people who have difficulties with typing.
The integration could be as simple as a microphone button next to the input box, allowing users to toggle voice input when required. Many platforms (Google Docs, mobile keyboards, etc.) already provide similar functionality, and bringing it into Cursor would make the chat experience much smoother and accessible.
I believe this feature would significantly improve usability, accessibility, and productivity for many Cursor users.
Thanks,
Saurabh Dhir
New Title Suggestion: âRevolutionize Text Input with Whisper Dictationâ
Imagine hands-free text entry thatâs not only fast, but incredibly accurate. Whisperâs cutting-edge voice dictation technology, powered by OpenAI, offers a superior alternative to the standard built-in speech-to-text solutions. With its ability to handle complex vocabulary and provide greater accuracy, Whisper can truly transform the way we interact with our favorite editor. Iâm asking the Cursor team to consider integrating support for Whisper dictation.
Why this feature matters:
⢠Unmatched Accuracy: Whisper outperforms traditional dictation tools, significantly reducing the need for corrections.
⢠Flexibility and Choice: Users could either access OpenAIâs Whisper API or connect a locally hosted Whisper model, depending on their preferences and workflow.
What Iâm requesting:
-
Seamless integration of OpenAIâs Whisper API for top-tier dictation capabilities.
-
Support for locally hosted Whisper models, providing more options for privacy-conscious users.
Letâs bring unparalleled voice-to-text performance to our editor and make text entry effortless and precise!
Mockup
I just found the dictation tool in Cursor, and itâs not half bad.
cmd+opt+v
It works in a document, like .md or notes, but it doesnât work in the compose,r which it should just be added.
Use macwhisper
TON OF FEATURE REQUESTS COULD BE ADDED
TON OF STUFF THESE GUYS CAN BUILD
SO FAR THEYâVE ONLY BEEN ABLE TO COME UP WITH COMPOSER
HAHA
I didnât manage to find this feature in Cursor. Is it still there?
I have built my own integration, which allows local Whisper transcription as well as using Open AI and Grok APIs for online transcription. I will be releasing an update soon and can update here as well if youâre interested. It works pretty well on Mac M1 local!
Hey, this feature is not available in Cursor yet, but it might be added in the future. We are considering the implementation of new features that users request, so what you are doing here is very helpful.
4 Likes
I have made a big release for this VSCode extension and itâs working great for me with Cursor on a Mac M1.
It would be great to get some feedback from the Cursor team @deanrie!
1 Like
For local transcription only SoX and Docker is required to get started and with a Mac M1 the transcription speed is much faster than previous versions of the extension.
Support for remote transcription through the OpenAI and Groq APIs is also included in this release for those lacking the machine power to support local transcription.
Cursor should buy this from you already 