Skip to content
Audio

Caller emotion models

эмоции из телефонного звонка

1 min readupdated

Набор быстрых моделей для распознавания эмоций в телефонных звонках на английском и арабском от ozonetg. Каждая модель держит речевой энкодер замороженным и добавляет небольшую обученную голову на 55 эмоциональных атрибутов; фиксированная таблица превращает их в семь эмоций.

Ключевые особенности

  • Семь эмоций: angry, disgusted, fearful, happy, neutral, sad, surprised.
  • Несколько вариантов: Qwen3-ASR и Parakeet в сжатом ONNX с головой, ONNX-модель Horizon для мультиязычных слов, голова для Whisper large-v3.
  • Обучение без человеческих меток эмоций: голова предсказывает 55 атрибутов, таблица переводит их в классы.
  • Parakeet-вариант дообучен на реальной телефонной речи; исходный чекпойнт в форме NeMo тоже есть в репозитории.
  • Для qwen/parakeet/horizon нужны numpy и onnxruntime; Whisper-голова требует torch, transformers и GPU.
  • Вход — mono float32, 16 kHz; scores() даёт семь оценок в порядке model.classes.

Пример

import sys
sys.path.insert(0, "models/qwen_emotion")
from qwen_emotion import QwenEmotion

model = QwenEmotion("models/qwen_emotion", threads=4)
scores = model.scores(pcm_float32_16k)
# семь оценок в порядке model.classes
ПодробнееИсследовательские результаты, на живых звонках не тестировались.
scores не калиброваны: порог алерта подбирайте по своему трафику.
Определять одну главную эмоцию на реплику пока слабо; для fearful и disgusted мало тестовых примеров.
Лицензии компонентов разные — проверяйте условия каждого источника.
Discuss Telegram 0 120

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.5 GB · ready files1.1 GB
  • FP8 / INT8recommendedweights 0.3 GB0.8 GB
  • 4-bitweights 0.1 GB0.7 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.

    # ozonetg/caller-emotion-public
    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 ozonetg/caller-emotion-public --local-dir ~/models/caller-emotion-public
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to caller-emotion-public
  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 caller-emotion-public (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ozonetg/caller-emotion-public
Author
ozonetg
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

1200

Ask AI about this model

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

Comments

Language models

Emotion-S1

Emotion-S1