# machine-learning
- p
PixPrucer
09/28/2022, 5:22 PMIt's okay Stephan will eventually be famous - u
{K EY1} (Kei)
09/28/2022, 5:22 PMPolish basemodel will be great - h
HolyArapaima
09/28/2022, 5:22 PMI should have sent one of my messages as a reply - h
HolyArapaima
09/28/2022, 5:23 PM@fatherallah - h
HolyArapaima
09/28/2022, 5:31 PMAnd if you don't feel a recording is helping your dataset, I just wouldn't add it, I only add things to my dataset that I am confident enough are gonna actually be beneficial because if it's something that it's really is gonna struggle in understanding it's just gonna blur your results in the end. - p
postmates!!
09/28/2022, 6:15 PMWHY DOES MY SPAMTON MODEL SOUND LIKE THAT 💀💀💀💀 - f
fatherallah
09/28/2022, 6:15 PMThanks for the advice i’ll take this away. Didn’t know that you sampled yourself singing. I’m trying to replicate your results with the acapellas of an artist. It should be the same method technically, but getting 45 minutes is difficult. - h
HolyArapaima
09/28/2022, 6:26 PMYeah the more word coverage and stuff the better if it's good data if it is not sometimes in can cloud the model - p
PixPrucer
09/28/2022, 6:27 PMIt's literally a fart - h
HolyArapaima
09/28/2022, 6:28 PMEspecially if there are big frequencies variants and factors that make the samples sound different then each other - h
HolyArapaima
09/28/2022, 6:28 PMEQ matching and stuff though can really help - p
postmates!!
09/28/2022, 6:33 PMsounds like something spamton would do tbf - h
hecko
09/28/2022, 7:38 PMfunny whisper fails - j
Justin
09/28/2022, 7:40 PM - a
a duck
09/28/2022, 9:24 PMcurrently sounds nothing like joe hawley - u
{K EY1} (Kei)
09/29/2022, 2:09 AMso i'm pretty new to ai image generation, but i'm using it for inspiration for a drawing rn and i want to get as many samples as possible what are some systems yalls would recommend? i tried dream/wombo ai, dall-e mini / craiyon, disco diffusion, and stable diffusion - u
{K EY1} (Kei)
09/29/2022, 2:11 AMi actually got some pretty cool results with dreaml- 2
- 1
- l
Lexi (delulu posts on the daily)
09/29/2022, 2:12 AMwow so magical - a
a duck
09/29/2022, 4:43 AMretrained the joe hawley model - h
hecko
09/29/2022, 9:10 AMfor generating concrete objects, stable diffusion no contest (except maybe disco diffusion which i haven't tried) - u
{K EY1} (Kei)
09/29/2022, 12:34 PMOo ok Thank you - m
mega b
09/29/2022, 2:33 PM - h
hecko
09/29/2022, 2:52 PMno way - h
hecko
09/29/2022, 2:52 PM@Pikachu ✓ - p
Pikachu ✓
09/29/2022, 2:54 PMOmg - h
hecko
09/29/2022, 2:56 PMquestion is, can it run on a t4 - p
Pikachu ✓
09/29/2022, 3:00 PMNo idea - p
postmates!!
09/29/2022, 7:32 PMif you wanna write a year like “2005” should you do it like that, or like “two thousand five” - u
{K EY1} (Kei)
09/29/2022, 7:35 PMEither way is fine since it converts numbers but just in case i always write it as the words