doc has a supir installer
doc has a supir installer
Ayou can also use seedvr2
JThank will check that
Dseems so
DI also have another question.. the answer migth be somewhere but i havent come across it yet.. in your video you said one can have as many consepts as theyd like
1. can i have a consept for the face like in your video with the masks and have another consept to do the whole body inclusive of face..
2. the lora is about 6gb so is it loaded as a checkpoint or using a lora loader..
1. can i have a consept for the face like in your video with the masks and have another consept to do the whole body inclusive of face..
2. the lora is about 6gb so is it loaded as a checkpoint or using a lora loader..
Kguys if i need to stop my lora training, i just need to open the last saved lora as the model later?
Ji think you need to set networklora or network*** something option
Jdont fix the skin tone
Awhat do you mean skin tone
Afrom black man /woman to white?
Jthats racist 
K@Furkan Gözükara SECourses ill ask the expert.
Dcan someone take a look at my dataset images and tell me if theyre good or not
.. im training for SDXL and im running mad why my training isnt even close to okay..im using one trainer but none of my trainings have been good
.. im training for SDXL and im running mad why my training isnt even close to okay..im using one trainer but none of my trainings have been good
CWhat you training ? Wan?, FLux, Qwen ?
Dsdxl....
Coh never trained.. so dont know what is best dataset
Ddamn...ive been looking..at the dataset in the Drs videos and when i look at mine.. it seems okay
CMaybe captions
Jwhich dataset where? which captions?
Factually on the contrary you can have single concept. from concept we mean single subject
Flike a person
Fnot 2 person
For an item
Fnot 2 item
Fyes you can provide this
F
Dand when it comes to the lora strength... what do you recomend.. assuming i did the exact training as your sdxl video on OT
J@Dr. Furkan Gözükara ^ spam
Aomg elon musk
Ahi
CHi!
I’m preparing for fine-tuning the F5 TTS model for Polish, since there isn’t one available yet. I did find one person who created a Polish model using about 90 hours of recordings and trained it on an A100 80GB for around 24 hours. Unfortunately, he didn’t share that model.
https://www.youtube.com/watch?v=K6vY9Je4ufQ
That’s why I decided to give it a try myself. There isn’t much information online about TTS training configurations, unlike with photo or video models. Based on what I managed to gather so far:
My dataset contains 142 hours of correct Polish speech. The dataset has been split into smaller files with transcripts (the transcription process is still ongoing).
As for the configuration, I’m not entirely sure if it’s correct, but I plan to start training with the following settings:
I don’t know if it will work, and I also don’t know how long it will take on an RTX 4090. Possibly a few days! XD
So, if anyone here has done a similar training and could help me out with tips or suggestions, I’d really appreciate it.
Yesterday, I ran a very short test training with just a 2-hour dataset. Unfortunately, the process crashed during the night, but it managed to reach 2500 steps. I saved sample outputs every 500 steps, so I have five of them.
I must say, at 500 steps the difference between the reference wav and the generated file was huge – as a native Polish speaker, I couldn’t understand a single word from the generated one. But at 2500 steps, it was already intelligible. Lots of mistakes, but at least I could understand the speech.
I could share the 2500-step sample here, but since it’s in Polish, I’m not sure if any of you would understand it.
Anyway, if someone can help, I’d be very grateful for any advice.
I’m preparing for fine-tuning the F5 TTS model for Polish, since there isn’t one available yet. I did find one person who created a Polish model using about 90 hours of recordings and trained it on an A100 80GB for around 24 hours. Unfortunately, he didn’t share that model.
https://www.youtube.com/watch?v=K6vY9Je4ufQ
That’s why I decided to give it a try myself. There isn’t much information online about TTS training configurations, unlike with photo or video models. Based on what I managed to gather so far:
My dataset contains 142 hours of correct Polish speech. The dataset has been split into smaller files with transcripts (the transcription process is still ongoing).
As for the configuration, I’m not entirely sure if it’s correct, but I plan to start training with the following settings:
I don’t know if it will work, and I also don’t know how long it will take on an RTX 4090. Possibly a few days! XD
So, if anyone here has done a similar training and could help me out with tips or suggestions, I’d really appreciate it.
Yesterday, I ran a very short test training with just a 2-hour dataset. Unfortunately, the process crashed during the night, but it managed to reach 2500 steps. I saved sample outputs every 500 steps, so I have five of them.
I must say, at 500 steps the difference between the reference wav and the generated file was huge – as a native Polish speaker, I couldn’t understand a single word from the generated one. But at 2500 steps, it was already intelligible. Lots of mistakes, but at least I could understand the speech.
I could share the 2500-step sample here, but since it’s in Polish, I’m not sure if any of you would understand it.
Anyway, if someone can help, I’d be very grateful for any advice.
Awhen training audio
Ayou need to remove as much noise as possible
Ato make it easier for the GAN
Amvsep.com is a great start
Aalso, resample the file to 32khz
Jvocoder:
type: vocos
config:
model_name: "charactr/vocos-mel-24khz"
Jmodel is 24khz?
�guys, anyone have a comfyui solution to the shiny skin problem? example here. i want skin without shine
Ait has to do with the lightning
Aalso try 2.8 guidance
�well this was result of a flux kontext face swap workflow
TThis is a feature of flux kontext, no way to avoid it completly, only post-process fix / gen over
TI was never able to reploduce amazing results using flux kontext workflow. Everything it generates looks like flux schnell - old generation / midjourney / sdxl quality
TIt excels at not touching parts of the image it doesn;'t need to touch, but for parts that is does - it woudl go full schnell bs
TMaybe it's better with FLUX.1 Kontext Pro version
Jfor 32gb vram should i use the 24gb one?
Jwhat is this?
Jfor DreamBooth configration
Jprobably
