Whisper-cpp using a small or medium English model may be a better fit for you than passing your audio through a general purpose LLM.
https://github.com/ggml-org/whisper.cpp
When that is done, an LLM can remove all your filler words and produce a cleaned up version.
My note was mostly about the voice of a writer. That when you write, your voice is captured in the words and grammar and pacing. LLMs produce a voice that is very recognizable. You can tell when a piece of writing has been promoted into existence and published without revision.
Jay
npub10m…x27hf
2026-05-07 13:51:34 UTC
in reply to nevent1q…y3lz
Author Public Key
npub10mtatsat7ph6rsq0w8u8npt8d86x4jfr2nqjnvld2439q6f8ugqq0x27hfPublished at
2026-05-07 13:51:34 UTCEvent JSON
{
"id": "2d8167459b85bef3882a0a8aad9fbd9bc787a80cd51635e6529864700b79d002",
"pubkey": "7ed7d5c3abf06fa1c00f71f879856769f46ac92354c129b3ed5562506927e200",
"created_at": 1778161894,
"kind": 1,
"tags": [
[
"e",
"cc3dd22534e50c7558d1b3c4b0dfbd8cf1d15ddadad0391f946f5c279f2085e5",
"",
"root"
],
[
"e",
"8b66ad6d3e6fa9ac4ca6bb9512b7c5c232a178e7da24738e13b9cf0decad3a69",
"",
"reply"
],
[
"p",
"7ed7d5c3abf06fa1c00f71f879856769f46ac92354c129b3ed5562506927e200"
],
[
"p",
"df478568479de26b4a83c1bdc4dbab61b5cc82e1a312e2b28bc815a12a951e67"
],
[
"r",
"https://github.com/ggml-org/whisper.cpp"
]
],
"content": "Whisper-cpp using a small or medium English model may be a better fit for you than passing your audio through a general purpose LLM.\n\nhttps://github.com/ggml-org/whisper.cpp\n\nWhen that is done, an LLM can remove all your filler words and produce a cleaned up version.\n\nMy note was mostly about the voice of a writer. That when you write, your voice is captured in the words and grammar and pacing. LLMs produce a voice that is very recognizable. You can tell when a piece of writing has been promoted into existence and published without revision.",
"sig": "e6e14f350eb4f3ea940a667582d5df555e6aac12e76179e3377ca8fa25d0712db9f3747c31c661dbaae1f3787d7e1cbe1842c689f1c59ccb2db3233258ad3983"
}