Local Image Generator โ€” model assets

Asset repository for the Obsidian plugin Local Image Generator (https://github.com/johannes-kaindl/local-image-generator). The plugin's built-in engine downloads these files only after the user clicks "Download", verifies each file against a SHA-256 pinned in the plugin release, and stores them outside the vault.

Contents

Path What it is
sd-turbo/text_encoder/model.onnx CLIP text encoder, fp16 weights, fp32 inputs/outputs
sd-turbo/unet/model.onnx UNet, fp16 weights, fp32 inputs/outputs (timestep is a 0-d float32 scalar)
sd-turbo/vae_decoder/model.onnx VAE decoder, fp16 weights, fp32 inputs/outputs
sd-turbo/tokenizer/vocab.json, merges.txt CLIP BPE tokenizer data
sd-turbo/LICENSE.md, sd-turbo/NOTICE.md Stability AI Community License and attribution
runtime/ort-<version>/ort-wasm-simd-threaded.asyncify.wasm ONNX Runtime Web (MIT), the exact build the plugin version links against

Provenance

The ONNX files are the plugin's own conversion of the official stabilityai/sd-turbo weights: optimum ONNX export, then onnxruntime's float16 conversion with keep_io_types=True. The script is tools/convert/convert_sd_turbo.py in the plugin repository; no third-party conversion is used. Powered by Stability AI.

License

The model files are redistributed under the Stability AI Community License (research, non-commercial and limited commercial use free of charge โ€” see the license for the revenue threshold). The ONNX Runtime WASM is MIT. The plugin itself is AGPL-3.0-or-later.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for v6t2b9/local-image-generator-models

Quantized
(8)
this model