You can deploy this model and start earning money today!

mlx-community/
AnimeGen-T2V-A14B-Lightning-bf16

Name: AnimeGen-T2V-A14B-Lightning-bf16
Availability: InStock

Wan-AI/Wan2.2-T2V-A14B

finetune

apache-2.0

Views

No views yet

AnimeGen-T2V-A14B-Lightning-bf16 (MLX)

An MLX (Apple Silicon) conversion of aidealab/AnimeGen-T2V — an anime-specialized full fine-tune of Wan2.2-T2V-A14B (dual-expert MoE) — in bf16, shipped as a pre-merged Lightning checkpoint for few-step, CFG-free inference.

Built for the Swift-MLX wan-core / Bernini-R runtime: the flat, 1095-key A14B checkpoint layout that runtime loads. For lighter footprint see the int4 variant.

What's in here

File	What
`high_noise_model.safetensors`	fine-tuned high-noise expert (bf16), Lightning-merged
`low_noise_model.safetensors`	fine-tuned low-noise expert (bf16), Lightning-merged
`vae.safetensors`	16-channel Wan VAE (stock Wan2.2)
`t5_encoder.safetensors`	umT5-XXL text encoder (stock Wan2.2)
`config.json`	Wan2.2-A14B config

How it was made

The two fine-tuned experts from aidealab/AnimeGen-T2V (original-Wan key naming, bf16) were converted 1:1 into the MLX A14B layout (sanitize_wan_transformer_weights; key set asserted equal to the base A14B contract — 1095 keys, 0 added / 0 dropped).
The lightx2v Wan2.2-Lightning Wan2.2-T2V-A14B-4steps-lora-250928 LoRA was merged offline per-expert with AIdeaLab's weighting (set_adapters([high, low], [2.0, 1.0]); kohya W += strength·(alpha/rank)·(up @ down)).

Recommended inference recipe (from the AnimeGen model card)

Sampler: euler, CFG-free (guidance 1.0)
8 steps, shift 3.0 (FlowMatchEuler)
832×480 or 1280×720, frames 16·seconds + 1 (must be 4n+1), 16 fps

License & attribution

Apache-2.0. Derived from aidealab/AnimeGen-T2V (AIdeaLab, developed with GENIAC/METI support) and the lightx2v Lightning LoRA. Unofficial community MLX conversion; not affiliated with AIdeaLab.

mlx-community/AnimeGen-T2V-A14B-Lightning-bf16