Modelpublic

SceneWorks/Lens-Turbo

sourceHugging Facemitupdated 3mo agoView on Hugging Face

0likes3downloads

Model Card

Lens-Turbo

Self-contained diffusers-layout snapshot of Microsoft's Lens-Turbo (distilled) text-to-image model, re-assembled for in-house (SceneWorks) use after Microsoft removed the original microsoft/Lens-Turbo repository from the Hub.

This is a repackage, not a retrain — every weight is byte-identical to a public upstream source (verified by tensor-level comparison against Comfy-Org's authentic redistribution):

Component	Source	License
`transformer/`	Lens-Turbo DiT, bf16 — from `Comfy-Org/Lens` (`diffusion_models/lens_turbo_bf16.safetensors`)	MIT
`text_encoder/`	gpt-oss-20b (MXFP4), used encoder-only — from `openai/gpt-oss-20b`	Apache-2.0
`tokenizer/`	gpt-oss tokenizer — from `openai/gpt-oss-20b`	Apache-2.0
`vae/`	FLUX.2 VAE (`AutoencoderKLFlux2`) — from `black-forest-labs/FLUX.2-dev`	FLUX.2-dev license

The text encoder is stock, frozen gpt-oss-20b and the VAE is stock FLUX.2-dev; only the transformer/ DiT is Lens-specific. The encoder + VAE are identical to those in SceneWorks/Lens — only the distilled transformer/ differs.

Layout

tokenizer/       tokenizer.json, tokenizer_config.json, special_tokens_map.json
text_encoder/    model-0000*-of-00002.safetensors (MXFP4), model.safetensors.index.json, config.json
transformer/     lens_turbo_bf16.safetensors
vae/             diffusion_pytorch_model.safetensors, config.json

Sampling defaults for the distilled model: 4 steps, guidance 1.0 (≈ no CFG).