mangalarora/moderation-models

Prixaro on-device moderation models

LiteRT models used by the Prixaro moderation SDK. All inference runs on the user's device.

The SDK reads the signed manifest at manifest/v1/manifest.json (Ed25519, detached signature manifest/v1/manifest.json.sig) and downloads files from the immutable commit named in its baseUrl. Every file is pinned by size and SHA-256; converted models passed a numeric parity check against their source graph.

Model	Version	Capability	Variants	Licence	Source
nudenet-320n	3.4.0	video.nsfw	fp16, dynamic_int8	LicenseRef-NudeNet-Commercial	https://api.github.com/repos/notAI-tech/NudeNet/releases/assets/176831997 @ immutable-ur
nudenet-640m	3.4.0	video.nsfw	fp16, dynamic_int8	LicenseRef-NudeNet-Commercial	https://api.github.com/repos/notAI-tech/NudeNet/releases/assets/176832019 @ immutable-ur
falcons-nsfw-vit	1.0.0	video.nsfw	weight_only_int8, dynamic_int8	Apache-2.0	hf://Falconsai/nsfw_image_detection @ 96cb0d0342c7
mivolo-v2	2.0.0	video.age_gender	weight_only_int8, dynamic_int8	Apache-2.0	hf://iitolstykh/mivolo_v2 @ 53393526c220
yunet-face	2023.3.0	video.face	fp16	BSD-3-Clause	hf://litert-community/YuNet-Face-LiteRT @ f08a6874cf8a
minifasnet-v2-liveness	1.0.0	video.liveness	fp32	Apache-2.0	hf://litert-community/Silent-Face-Anti-Spoofing-LiteRT @ a93eeab801f4
ppocr-v5	5.0.0	video.ocr	fp16	Apache-2.0	hf://litert-community/PP-OCRv5-LiteRT @ cc63dcd9fa8d
silero-vad	5.0.0	audio.transcription	fp32	MIT	hf://soniqo/Silero-VAD-v5-LiteRT @ 655bff6b9a74
whisper-tiny	1.0.0	audio.language_id	dynamic_int8	Apache-2.0	hf://litert-community/whisper-tiny @ d92f0ee1d431
fasttext-lid176	1.0.0	text.abuse	native	CC-BY-SA-3.0	https://dl.fbaipublicfiles.com/fasttext/supervised-models/lid.176.ftz @ immutable-ur

Each model is subject to its own licence listed above; notes are recorded per model in the manifest.