Skip to content
Other

GeLDA checkpoints for speech emotion recognition

1 min read

The GeLDA collection of checkpoints implements semantically aware generative augmentation of latent data (hidden numerical representations) for zero-shot speech emotion recognition (recognizing emotions in speech without prior training on the target language). Created by researchers Jaesung Bae and Minje Kim, the model improves emotional tone detection in voices using powerful audio models like Whisper and WavLM. The solution relies on latent diffusion models and specialized task adapters. This set is useful for speech technology researchers and voice analysis system developers.

Key features

  • Support for popular base audio models, including whisper-large-v3, wavlm-large, and emotion2vec-base.
  • Solutions for zero-shot and few-shot emotion recognition scenarios across multiple languages, including Russian.
  • Application of latent diffusion in various spaces Z(0), Z(1), and Z(2) for feature generation.
  • Ready-to-use full task adapters (Stage-3 task adapters) trained using GeLDA.

What you can do

  • Run latent diffusion model inference to generate synthetic emotional features.
  • Integrate adapter checkpoints into audio message emotion analysis pipelines.
  • Conduct research and experiments on zero-shot speech recognition using Russian-language datasets.

Example

from gelda.diffusion import load_inference_checkpoint
denoiser, config = load_inference_checkpoint("zero_shot/whisper-large-v3/french/fold1/diffusion_z0.safetensors")
Read moreDetails
Zero-shot setups are supported for various language groups, including Russian with whisper-large-v3.
Folds are structured similarly to the EmoBox framework.
The checkpoints correspond to repetition 0 from the authors' research paper.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # jaesungbae/GeLDA
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download jaesungbae/GeLDA --local-dir ~/models/GeLDA
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to GeLDA
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of GeLDA (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

jaesungbae/GeLDA
Author
jaesungbae
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments