>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?

>>105649696
>different parameter count
>that's my point, SDXL is a 2.7b parameter model, Chroma is a 8.9b parameter model, if a 2.7 parameter model was able to eat so much anime knowledge, a 8.9b one definitely can too

I'm a fucking noob , but as far as i've understood,

number of parameters of a model only affect HOW the INITIAL data is transformed in a meaningful output.

So it all depends on the initial quantity and quality of the training picture database?