+- Like Ra's Naughty Forum (https://www.likera.com/forum/mybb)
+-- Forum: Fetishes, obsessions, traits, features, peculiarities (https://www.likera.com/forum/mybb/Forum-Fetishes-obsessions-traits-features-peculiarities)
+--- Forum: Hypnosis and Hypno-fetish (https://www.likera.com/forum/mybb/Forum-Hypnosis-and-Hypno-fetish)
+--- Thread: Hypnosis Scripts (/Thread-Hypnosis-Scripts)
RE: Hypnosis Scripts - Like Ra - 26 May 2026
As a part of my Hermes agentic AI project (https://www.likera.com/forum/mybb/showthread.php?tid=4606), I started automatically transcribe all hypno files I have. It might take months (if not years), but it should be done, right? And since there are some obvious errors, all TXT files have to be looked at manually.
As a test I auto transcribed all Goddess Gracie (because there is only one voice, what means less errors) files I have (~200). While doing quick fixes it can be clearly seen, that GG should be avoided š
Most likely, Nikki Fatale will be the next. At the moment I would avoid multichannel/multivoiced/layered files. Any suggestions?
The ultimate idea is to train a hypno-LoRA.
RE: Hypnosis Scripts - AI_addict_hypnofan - 26 May 2026
Elevenlabs' Scribe v2 is much faster and higher transcription accuracy than Whisper.
Just deposit $10 to Elevenlabs or another site that offers it, and get your API saved on your computer.
Then ask your agentic AI (e.g. Hermes, Codex, Grok Build) to send all the files in a directory over to it for transcription, one by one.
It's pretty quick - 200 files would be an afternoon I guess, with no involvement needed from you.
After that, you can get the agentic AI to go through them checking for any punctuation and formatting errors, and perhaps to edit broken sentences and ums-and-ahs. Well, actually you can tell it to do this in your initial prompt, so you'll come back to all the transcriptions finished and tidied.
RE: Hypnosis Scripts - Like Ra - 27 May 2026
(26 May 2026, 20:23 )AI_addict_hypnofan Wrote: 200 files
(26 May 2026, 20:23 )AI_addict_hypnofan Wrote: Just deposit $10
I've got 25k files.
RE: Hypnosis Scripts - subarashi - 09 Jul 2026
(26 May 2026, 16:45 )LikeĀ Ra Wrote: As a part of my Hermes agentic AI project (https://www.likera.com/forum/mybb/showthread.php?tid=4606), I started automatically transcribe all hypno files I have. It might take months (if not years), but it should be done, right? And since there are some obvious errors, all TXT files have to be looked at manually.
As a test I auto transcribed all Goddess GracieĀ (because there is only one voice, what means less errors) files I have (~200). While doing quick fixes it can be clearly seen, that GG should be avoided š
Most likely, Nikki Fatale will be the next. At the moment I would avoid multichannel/multivoiced/layered files. Any suggestions?
The ultimate idea is to train a hypno-LoRA.
Nikki Fatale, the tale says legends neverĀ die.
I probably be asking for a ban, but was looking for these in long time (some were never publicly for sale):
Confidence with Women
Dominant Man
Hypnotic Playtime
Meme Filtered
Renew Red -Ā Underground version
Valentines / Valentines Special
---
Anyway,
about your idea for LoRA, it requires question-response pairs to be effective. You could probably synthesise these by summarising the translated scripts and use these as questions.
For script generation, it doesn't really matter much - simply concatenating random chunks from multiple scripts can work as an initial prompt very well, but you need to reach ~100K tokens in size to shift token probability in your favor (assuming model context-window ~1M). For example in Bambi Sleep case around 30% of the original scripts are needed to reach critical mass and flip the model into Bambi Writer.
These files use many tricks, so adding some books on psychology and hypnosis into the mix could probably yield better results.
Don't worry about translation errors - they don't matter at all. The LLM gets context from longer chunks, so single-word bumps don't affect the result in any meaningful way.