Feature Requests

Allow users to define/use language models other than the providers ( imagine cohere, RAG api, etc )

Could we add an option under AI Models > Language Models to select a custom OpenAI model, instead of being restricted to the dropdown options? For instance, I’ve created an assistant at https://platform.openai.com/assistants , and I’d like to have that assistant handle the requests instead of the default OpenAI models. If there’s already a way to do this, please let me know!

Was playing around with language models but i could not chose the could ones (don't know why) and tried the local language model but it was to slow. I deleted it again but now i can't do any recording since it says i don't have a language model? No way to go back to not having a language model.

Right now, the biggest limiting factor in using the iOS app more is the slowness (and typos) with which the voice model transcribes audio to text. Given I have an iPhone 15 Pro, which is at the high end of processing speed relative to other users, it's still quite slow. It would be fantastic if we could use our own API key to access Whisper or, potentially, DeepGram as well. This would significantly boost my usage of the iOS app, as right now, it is sluggish and has some transcription errors when using either the Ultra or Ultra Light models. Thank you very much for your consideration.

Include models fine tuned specifically for Spanish speech.

Hi, I'm moving from Murmurtype to Superwhisper, but as I was doing some testing, I found out on my Mac M1 Pro, that Murmurtype returns result significantly faster. Enough for me to keep using MMT and run Superwhisper in parallel for situations where MMT fails to transcribe (cause MMT fails very ungracefully). If I could have an option for using a mode that sends data for transcription to cloud like Murmur Type does, then I would definitely move to Super Whisper only. And I would aim for local transcription when recording more private matters or when the cloud would fail and I would use the local one as a plan B solution.

Would it be possible to add support for custom voice models, especially cloud models? It would be awesome if we could use an OpenAI API token to access the Whisper API. It's super fast and accurate – I use it with a Chrome extension for ChatGPT, and it's fantastic. One more thing: adding the ability to provide customers the option to use external locally hosted voice models, deployed via container app for example. https://github.com/ahmetoner/whisper-asr-webservice I think it would result in much more accurate transcription, with top model accuracy running on a local GPU with the benifit of being complely private and off the cloud. Thanks for considering this! I think it would take SuperWhisper to the next level.

The open-source iOS app WhisperBoard offers the Large v2 and v3 models for all languages, and it works much better on my iPhone 13 Pro than superwhisper (I'm French). Please make these models available on the iOS app.

Include models fine tuned for Cantonese speech

Currently, the literal punctuation option only supports dictated punctuation in English. It would be great if we could get this feature in other languages as well.