Feed the model
Data & Memory
Add text for emotion learning. Upload images — the model stores them in weights and learns to reconstruct them.
Text Example
Image — Visual Memory
Dataset
Nothing yet.
Emotion classifier
Text Model
Bag-of-words + bigrams trained with gradient descent. Any language.
Parameters
Not trained yet.
Encoder · Decoder · Backprop
Image Memory
Real neural autoencoder. Upload images → train → watch the model remember them in real time.
Architecture
ENCODER 32×32×3 → 128 → 32 latent
DECODER 32 latent + 62 emotions → 128 → 32×32×3
Loss: MSE · Optimizer: SGD · ReLU + Sigmoid
DECODER 32 latent + 62 emotions → 128 → 32×32×3
Loss: MSE · Optimizer: SGD · ReLU + Sigmoid
Train
Upload images first.
Memory
Left = original · Right = what the model remembers · Updates live.
What does the model feel?
Predict
Talk to your model
Chat
Let the model write
Generate
Select Emotions
Save everything, everywhere
Server & Sync
Your model saves automatically to your DomainNesia server. Never lose progress.
Server Configuration
Upload model_save.php and model_load.php to your DomainNesia hosting.
Then paste the URL and secret below. That's it.
Then paste the URL and secret below. That's it.
Sync Status
Last saved to serverNever
Last loaded from serverNever
Auto-saveAfter every train
Local Backup (fallback)
Download the full model as JSON — use as emergency backup if server is down.
Quick Load
Danger