Update ASR owner licensing and concise speech bundle documentation

a91a6c0 verified 2 days ago

1.69 kB

language:
  - kk
  - ru
  - en
license: other
license_name: nurgali-kadyrbek-asr-noncommercial-1.0
license_link: https://huggingface.co/farabi-lab/farabi-speech-models/blob/main/LICENSE
extra_gated_prompt: >-
  The original ASR contributions require a separate prior written agreement for
  commercial use. Read LICENSE and LICENSES.md for scope and component terms.
extra_gated_fields:
  I accept the ASR non-commercial license and the retained third-party component terms: checkbox
tags:
  - automatic-speech-recognition
  - text-to-speech
  - offline
  - model-bundle

Al-Farabi speech models

Offline installation bundle for the Al-Farabi Assistant:

ASR: nur-dev/asr-220m-kk-ru-en — Kazakh, Russian, English and mixed Kazakh–Russian recognition; 16 kHz input.
TTS: nur-dev/ait-syn-omnivoice-72-voices — 72 prepared voices; 24 kHz output, default voice kk_03.

Includes weights, runtime code, tokenizers, voice presets, component notices and resolved dependencies. It is an installation archive, not a Transformers checkpoint. Exact revisions and checksums are in speech-models.manifest.json.

hf download farabi-lab/farabi-speech-models --local-dir speech-bundle
python3 speech-bundle/ops/speech_models.py install speech-bundle/farabi-speech-models.zip --root .
./farabi-app --open

Provision compatible Python/CUDA environments before offline operation.

ASR original adaptations and original code: copyright Nurgali Kadyrbek. Non-commercial use only; commercial use requires a separate prior written agreement. See LICENSE. TTS and third-party components retain their respective terms in LICENSES.md and the included notices.