AIFX VOICE

Neural Training Workspace

Deploy high-fidelity voice models by feeding acoustic training inputs, configuring training epochs, and orchestrating your local GPU backpropagation runs.

Training Settings

Use lowercase, hyphens, and no special characters.
A5000 Limit

Pipeline Flowchart Map

Phase 1IDLE
Dataset SyncSequence uploading and streaming clean vocal cut files to Cloudflare R2 bucket.
Phase 2IDLE
Queue TelemetryCentral database job registry, scheduling local GPU hardware thread.
Phase 3IDLE
Acoustic PreprocessingLocal Applio resampler cleaning audio clips, splicing silence nodes.
Phase 4IDLE
Feature ExtractionEncoding and extracting base F0 pitch files using RMVPE algorithm.
Phase 5IDLE
Neural BackpropagationCore deep learning epoch cycles optimizing target voice parameters.
Phase 6IDLE
Sync weightsCompiling trained weight formats (.pth + .index) back to central storage.

Local RTX A5000 Hardware Telemetry

GPU Core Load2%
VRAM Allocated1.2 GB
GPU Core Temp41°C
Fan Speed
22%

Active Backpropagation Loss curve

Generator Loss Pitch Loss
1.50.80.0Ep 0Ep 75Ep 150
Simulated EpochIdle
Generator Loss0.00
Discriminator Loss0.00
Learning Rate0.00

Training History & Queue

Connecting to training logs database...