Published by TokForge for on-device inference.

TokForge

Website: https://tokforge.ai
Discord: https://discord.gg/Acv3CBtfVm
Google Play: https://play.google.com/store/apps/details?id=dev.tokforge
iOS TestFlight: https://testflight.apple.com/join/jnufjzRr

Runs on-device in the TokForge app.

TokForge: RealVisXL V4.0 Lightning (Qualcomm Hexagon NPU)

RealVisXL V4.0 Lightning image generation for the Qualcomm Hexagon NPU (HTP), packaged for on-device image generation in the TokForge Android app (dev.tokforge). This is the higher-quality 1024×1024 "Faithful" tier alongside the SD1.5 NPU bundle.

vc410: this repo replaces the previous SDXL-Turbo bins (few-step distillation gave fused heads / doubled bodies / extra limbs) with RealVisXL V4.0 Lightning, which is adversarially distilled for deterministic few-step Euler and renders clean, photoreal people. The W8A16 NPU quant is unchanged; the checkpoint was the cure.

The model is quantized to W8A16 (8-bit weights, 16-bit activations) and compiled to QNN HTP context binaries that run on the phone's Hexagon DSP. The pipeline uses fp16 text encoders and a TAESDXL tiny-VAE decoder, runs 6-step EulerDiscrete, and is guidance-free (one UNet pass per step).

Based on

SG161222/RealVisXL_V4.0_Lightning: CreativeML OpenRAIL++-M license, plus SG_161222's Civitai terms for RealVisXL (credit required, no selling the model or merges of it).

Format

QNN HTP context binaries (W8A16), native 1024px (128×128 latent). These are compiled for the Hexagon DSP and are not a portable format like GGUF. The repo ships native sets for V73 / V75 / V79 / V81; the app reads the device Hexagon arch (dsp_arch) and downloads the matching set. Forward-compatibility (a lower-arch binary also runs on a higher-arch DSP) still applies as a fallback. Device-verified clean on V75 (Lenovo SM8650) and V81 (RedMagic SM8850).

Pipeline

Stage	File
CLIP-L text encoder (fp16)	`<arch>/text_encoder_1_fp16.bin`
OpenCLIP-bigG text encoder (fp16)	`<arch>/text_encoder_2_fp16.bin`
Combined-emb MLP (host CPU)	`<arch>/sdxl_emb_mlp.bin`
UNet (W8A16, DSP)	`<arch>/unet.bin`
TAESDXL tiny-VAE (CPU MNN)	`<arch>/taesdxl_decoder.mnn`
Dual CLIP BPE tokenizers	`<arch>/tokenizer/`, `<arch>/tokenizer_2/`

Scheduler: EulerDiscrete (deterministic), trailing spacing, epsilon prediction, 6 steps, guidance_scale 0. Runs via the libsdxl_qnn_driver in the TokForge app.

See manifest.json for the per-arch file list, sizes, and md5 checksums.

Reference-image (IP-Adapter / identity-preserving): `ip/` subfolder

The ip/ subfolder ships the IP-Adapter context bins for identity-preserving (reference-image) rendering. When a reference photo is attached, the driver swaps the base unet.bin for the IP-baked UNet (unet_ip.bin) and renders with the reference's identity; no-reference renders are unchanged.

File	What
`ip/v75/unet_ip.bin`	IP-Adapter UNet context bin, Hexagon V75 (device-verified)
`ip/v81/unet_ip.bin`	IP-Adapter UNet context bin, Hexagon V81 (compiled from the same DLC; on-device validation pending)
`ip/model_w8a16.dlc`	arch-agnostic W8A16 QAIRT 2.40 DLC the per-arch `unet_ip.bin` are compiled from
`ip/ip_kv_proj_weights.npz`	70× `to_k_ip`/`to_v_ip` Linear weights for the CPU IP side (`k_0..k_69`, `v_0..v_69`, `inner[70]`, `heads[70]`)

The bins are identity-agnostic and safe to publish: the 70 cross-attention k_ip/v_ip are render-constant inputs computed on the CPU per render (reference → CLIP-ViT-H penultimate → plus-face Resampler → 70× to_k_ip/to_v_ip), so no likeness is baked into the published files. The CLIP-ViT-H encoder + plus-face adapter weights come from the standard public h94/IP-Adapter SDXL plus-face files, converted here for the NPU and CPU routes. The app downloads unet_ip.bin as an optional bundle file (paired in each set's files map); its absence degrades to a plain (no-identity) render and never fails the install. IO contract, checksums, and per-arch details are in manifest.json under the top-level ip block.

Credits

RealVisXL V4.0 Lightning by SG_161222 (Civitai, Hugging Face). Used under the CreativeML OpenRAIL++-M license and the creator's Civitai terms. SG_161222's Civitai terms: credit required, no selling the model or merges of it.
Stable Diffusion XL 1.0 by Stability AI (Hugging Face). Used under the CreativeML OpenRAIL++-M license.
IP-Adapter by Tencent AI Lab (h94) (Hugging Face). Used under the Apache-2.0 license.
OpenCLIP ViT-H-14 image encoder by LAION (Hugging Face). Used under the MIT license.
TAESDXL tiny VAE by madebyollin (Hugging Face). Used under the MIT license.
Compiled for the Hexagon NPU with Qualcomm AI Runtime (QAIRT) 2.40 by Qualcomm.
Converted to MNN with MNN by Alibaba (Apache-2.0).
The qnn_runtime/ files are Qualcomm AI Runtime (QAIRT) 2.40 runtime libraries by Qualcomm Technologies, Inc. They are Qualcomm software under Qualcomm's AI Stack License (the LICENSE.pdf shipped in the QAIRT SDK), not under the model license above, and are redistributed here for use inside the TokForge app only.