{"id":12617,"date":"2026-11-17T11:01:27","date_gmt":"2026-11-17T02:01:27","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=12617"},"modified":"2026-11-17T11:10:35","modified_gmt":"2026-11-17T02:10:35","slug":"lodestones-kromoss-open-weight-image-generation-model","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/lodestones-kromoss-open-weight-image-generation-model\/","title":{"rendered":"Kromoss 1024px Image Generation Model: 32GB+ VRAM, File List"},"content":{"rendered":"<p><em>Sample outputs are available on the <a href=\"https:\/\/huggingface.co\/lodestones\/Kromoss\">model card<\/a>.<\/em><\/p>\n<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/lodestones\/Kromoss\">lodestones\/Kromoss<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-openmoss-team-moss-image-en\/\">MOSS-Image guide (2 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-10-19<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>apache-2.0<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">safetensors<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Unverified (not confirmed by a primary source)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>It is reported that an open-weight model for generating high-quality images from text, &#8220;lodestones\/Kromoss&#8221;, has been released on Hugging Face. Please note, however, that this information is based on a public repository published by an independent developer, and is unconfirmed information for which official verification and announcements have not yet been completed.<\/p>\n<p>Kromoss is said to be a model that has undergone full fine-tuning at a resolution of 1024\u00d71024 pixels on the base model &#8220;MOSS-Image-Base&#8221; published by the OpenMOSS-Team. It is a text-to-image model that takes text prompts as input and outputs images, adopting the same architecture, inference pipeline, text encoder (Qwen3.5-9B), and VAE (FLUX.2 small decoder) as the preceding derivative version &#8220;MOSS-Image-SFT&#8221;. The repository structure is also designed to maintain compatibility, and it is explained that it can be used as a drop-in replacement for existing environments. It is reported that this fine-tuning focuses primarily on resolving the image blurriness that was considered an issue in the base model. Furthermore, a LoRA named &#8220;8-steps-dmd-lora.safetensors&#8221; created via data-space DMD (Distribution Matching Distillation) distillation, which enables high-speed generation in 8 steps, is provided at the same time.<\/p>\n<h2>Specifications<\/h2>\n<p>Specifications based on the published model card and repository information are as follows:<\/p>\n<ul>\n<li>Architecture: Transformer-based diffusion model. Adopts Qwen3.5-9B as the text encoder and FLUX.2 small decoder as the VAE decoder. As internal data types, bfloat16 is used for the Transformer and text encoder, while float32 is used for the VAE and initial noise generation.<\/li>\n<li>Output Specifications: Generation at 1024\u00d71024 pixels is standard, and resolution specification supporting multiples of 16 is available when calling from the Python API.<\/li>\n<li>Recommended Sampling Settings (Full Model Alone): Flow-matching Euler sampler, 50 steps, CFG (guidance scale) of 3.0, and Shift parameter of 5. As a layer skipping setting, skipping index 22 on the negative branch side is specified.<\/li>\n<li>Recommended Sampling Settings (When applying 8-step LoRA): Flow-matching Euler sampler (euler \/ simple settings for KSampler in ComfyUI), 8 steps, CFG of 1.0, and Shift parameter of 5. Since the negative branch is not used, layer skipping is set to none.<\/li>\n<li>High-speed LoRA (8-steps-dmd-lora.safetensors) Specifications: Composed of fp32 precision, rank 128, and alpha 128, targeting linear layers of each block (proj.q, k, v, o, ffn_up_gate, ffn_down). Because the weight differences are close to the bfloat16 rounding width, it is recommended to apply the weights unmerged rather than merging them directly into the base model weights.<\/li>\n<li>Training Specifications: Trained on 448,000 (448k) training images excluding those under 896\u00d7896 pixels, training all Transformer parameters in a 1024-pixel environment at a scale of 64 images per update step. The optimization method adopts the Muon-AdamW hybrid (learning rate 9e-6, warmup 200 steps), applying truncated logit-normal timesteps, q\/k normalization decay of the final block, and an EMA (Exponential Moving Average) coefficient of 0.999. The published weights are stated to be the EMA weights at the 7,000 update step.<\/li>\n<li>License: Specified as the Apache-2.0 license in the repository metadata.<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>Benchmark results presented by the publisher on the model card are as follows. A comparison table measuring CLIP ViT-L\/14 scores targeting PartiPrompts (100 prompts) has been published (figures for Kromoss are stated as measured values at update step 2000).<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Model<\/th>\n<th>CLIP<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>MOSS-Image-Base, 50 steps, CFG 3<\/td>\n<td>27.19<\/td>\n<\/tr>\n<tr>\n<td>MOSS-Image-Base + LoRA, 8 steps, CFG 1<\/td>\n<td>27.53<\/td>\n<\/tr>\n<tr>\n<td>Kromoss, 50 steps, CFG 3<\/td>\n<td>26.88<\/td>\n<\/tr>\n<tr>\n<td>Kromoss + LoRA, 8 steps, CFG 1<\/td>\n<td><strong>27.82<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>According to the published measurement results, the configuration combined with the 8-step high-speed LoRA (Kromoss + LoRA, 8 steps, CFG 1) achieves a score of 27.82, showing the highest numerical value, surpassing the base model alone or the base model with the same LoRA applied. On the other hand, Kromoss in standalone operation without LoRA applied (50 steps, CFG 3) stays at a score of 26.88, failing to reach the standalone score of the original MOSS-Image-Base model (27.19).<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>It is stated that this model was developed with the aim of improving image clarity while inheriting the text-to-image generation capabilities possessed by the base model, MOSS-Image-Base. As the performance and features of the original model, it supports image generation using natural language prompt inputs as well as tag enumerations and spatial layout prompts written in JSON format. With the JSON layout function, it is stated that by specifying bounding boxes (bbox_format: &#8220;yxyx&#8221;, bbox_coord_space: &#8220;abs_centralized&#8221;), the arrangement of the text to be drawn and the positional relationship of subjects can be structurally controlled in detail.<\/p>\n<p>As intended use cases specific to Kromoss, image generation at 1024\u00d71024 pixel resolution with reduced image blurriness\u2014which had been pointed out in MOSS-Image-Base\u2014can be mentioned. Because it shares the same file structure and inference configuration as the official derivative version MOSS-Image-SFT, it can be used for replacing existing SFT workflows.<\/p>\n<p>In addition, by utilizing the bundled distillation LoRA &#8220;8-steps-dmd-lora.safetensors&#8221;, it is designed to enable image generation with an extremely lightweight inference step count of just 8 steps and CFG 1. This is reported to make it suitable for scenarios where generation time is to be shortened while maintaining quality, or prototyping use cases for quickly verifying compositions.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 5.1B parameters (taken from the base model OpenMOSS-Team\/MOSS-Image-Base)<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>32GB (RTX 5090, etc.)<\/td>\n<td>BF16<\/td>\n<td>25.3GB<\/td>\n<td>30.4GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. This release is an adapter (LoRA etc.); the table shows what the base model <a href=\"https:\/\/huggingface.co\/OpenMOSS-Team\/MOSS-Image-Base\">OpenMOSS-Team\/MOSS-Image-Base<\/a> needs. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p>The publisher distributes this model as safetensors.<\/p>\n<p><strong>License \u2014 <code>apache-2.0<\/code> (Commercial use allowed):<\/strong> Permits commercial use, modification and redistribution. Redistribution requires including the license and stating changes; includes a patent grant.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats and the license field. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:peers --><\/p>\n<h2>Recent Models in the Same Size Class<\/h2>\n<p><em>Models with <\/em><em>4\u201315B<\/em><em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site&#8217;s estimates; licenses are as stated on the model cards.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<p class=\"lmw-table-hint\" style=\"margin:0 0 4px;font-size:0.85em;opacity:0.7;\">\u2192 Scroll horizontally to see all columns<\/p>\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Model<\/th>\n<th>Parameters<\/th>\n<th>Smallest VRAM tier<\/th>\n<th>License<\/th>\n<th>Our article<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Qwen\/Qwen-Image-2.1-Turbo<\/td>\n<td>7.1B<\/td>\n<td>48GB<\/td>\n<td>qwen-research<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/qwen-image-2-1-turbo-2\/\">Qwen-Image-2.1-Turbo 8-Step Fast Image Generation Model: 48GB+ VRAM<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>briaai\/Fibo-2.0-dev<\/td>\n<td>8.1B<\/td>\n<td>32GB<\/td>\n<td>bria-fibo<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/bria-ai-releases-fibo-2-0-dev\/\">Fibo-2.0-dev Image Generation Model: 32GB+ VRAM, File List<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>Viggle\/Qwen-Image-2.1-viggle-turbo<\/td>\n<td>7.1B<\/td>\n<td>48GB<\/td>\n<td>qwen-research<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/23\/qwen-image-2-1-viggle-turbo-v0-1-preview\/\">Qwen-Image-2.1-viggle-turbo Image Generation Model: 48GB+ VRAM<\/a> (2026-09-23)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:peers --><\/p>\n<p><!-- lmw:files --><\/p>\n<h2>Distributed Files<\/h2>\n<p><em>Weight files published in <a href=\"https:\/\/huggingface.co\/lodestones\/Kromoss\/tree\/main\">lodestones\/Kromoss<\/a>, listed by this site from the Hugging Face API. Sizes are the actual file sizes.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>File<\/th>\n<th>Size<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><code>8-steps-dmd-lora.safetensors<\/code><\/td>\n<td>1.19GB<\/td>\n<\/tr>\n<tr>\n<td><code>diffusion_models\/kromoss_bf16.safetensors<\/code><\/td>\n<td>10.16GB<\/td>\n<\/tr>\n<tr>\n<td><code>diffusion_models\/kromoss_int8_convrot.safetensors<\/code><\/td>\n<td>5.41GB<\/td>\n<\/tr>\n<tr>\n<td><code>text_encoder\/model-*-of-00004.safetensors<\/code><\/td>\n<td>16.79GB (4 split files)<\/td>\n<\/tr>\n<tr>\n<td><code>text_encoders\/qwen3.5_9b_bf16.safetensors<\/code><\/td>\n<td>19.31GB<\/td>\n<\/tr>\n<tr>\n<td><code>transformer\/diffusion_pytorch_model-*-of-00003.safetensors<\/code><\/td>\n<td>10.16GB (3 split files)<\/td>\n<\/tr>\n<tr>\n<td><code>vae\/diffusion_pytorch_model.safetensors<\/code><\/td>\n<td>250MB<\/td>\n<\/tr>\n<tr>\n<td><code>vae\/flux2-vae_small.safetensors<\/code><\/td>\n<td>250MB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:files --><\/p>\n<h2>How to Get It<\/h2>\n<p>Model weights and related files for lodestones\/Kromoss can be obtained from the Hugging Face repository. Because this repository is not a gated model, downloads are possible without performing prior application or consent procedures. The distribution format is mainly provided in safetensors format and can be used from Diffusers pipelines and ComfyUI environments.<\/p>\n<h3>Using with Diffusers<\/h3>\n<p>When performing inference using Diffusers, it is recommended to install the following packages in advance:<\/p>\n<pre><code class=\"language-bash\">pip install &quot;diffusers&gt;=0.38&quot; &quot;transformers&gt;=5.3&quot; &quot;accelerate&gt;=1.10&quot; &quot;safetensors&gt;=0.6&quot; &quot;Pillow&gt;=10&quot;\n<\/code><\/pre>\n<p>The basic calling procedure from Python code is as follows. Specify <code>pipeline_moss_image<\/code> as the custom pipeline, and load the Transformer and text encoder in bfloat16 and the VAE in float32.<\/p>\n<pre><code class=\"language-python\">import torch\nfrom diffusers import DiffusionPipeline\n\npipe = DiffusionPipeline.from_pretrained(\n    &quot;lodestones\/Kromoss&quot;,\n    custom_pipeline=&quot;pipeline_moss_image&quot;,\n    trust_remote_code=True,\n    torch_dtype={&quot;default&quot;: torch.bfloat16, &quot;vae&quot;: torch.float32},\n)\npipe.scheduler = type(pipe.scheduler).from_config(pipe.scheduler.config)\npipe.to(&quot;cuda&quot;)\n\nimage = pipe(\n    &quot;a tabby cat holding a sign that says KROMOSS&quot;,\n    height=1024,\n    width=1024,\n    num_inference_steps=50,\n    guidance_scale=3.0,\n    uncond_skip_layers=[22],\n    generator=torch.Generator(&quot;cpu&quot;).manual_seed(42),\n).images[0]\nimage.save(&quot;kromoss.png&quot;)\n<\/code><\/pre>\n<p>When using the 8-step LoRA, it is instructed to apply it via forward hooks rather than merging it directly into the base weights. Below is an example of the application code presented in the repository:<\/p>\n<pre><code class=\"language-python\">from safetensors.torch import load_file\n\ndef apply_lora(transformer, path, strength=1.0):\n    state = load_file(path, device=&quot;cuda&quot;)\n    modules = dict(transformer.named_modules())\n    for key in [k for k in state if k.endswith(&quot;.lora_down.weight&quot;)]:\n        name = key[len(&quot;diffusion_model.&quot;):-len(&quot;.lora_down.weight&quot;)]\n        down, up = state[key].float(), state[key.replace(&quot;lora_down&quot;, &quot;lora_up&quot;)].float()\n        scale = strength * state[f&quot;diffusion_model.{name}.alpha&quot;].item() \/ down.shape[0]\n        def hook(module, args, out, down=down, up=up, scale=scale):\n            return out + (scale * (args[0].float() @ down.T) @ up.T).to(out.dtype)\n        modules[name].register_forward_hook(hook)\n\napply_lora(pipe.transformer, &quot;8-steps-dmd-lora.safetensors&quot;)\nimage = pipe(\n    prompt,\n    height=1024,\n    width=1024,\n    num_inference_steps=8,\n    guidance_scale=1.0,\n    generator=torch.Generator(&quot;cpu&quot;).manual_seed(42),\n).images[0]\n<\/code><\/pre>\n<h3>Using with ComfyUI<\/h3>\n<p>The current ComfyUI master branch incorporates native support for MOSS-Image. Single-file format weights adopt the bf16 format with concatenated <code>proj.qkv<\/code>, similar to Comfy-Org&#8217;s distribution format. Place and load each component in the following directories:<\/p>\n<ul>\n<li><code>kromoss_bf16.safetensors<\/code> (or the quantized version <code>kromoss_int8_convrot.safetensors<\/code>): Place in <code>models\/diffusion_models\/<\/code> and load with the Load Diffusion Model node.<\/li>\n<li><code>qwen3.5_9b_bf16.safetensors<\/code>: Place in <code>models\/text_encoders\/<\/code> and load with the Load CLIP node by setting the type to <code>moss_image<\/code>.<\/li>\n<li><code>flux2-vae_small.safetensors<\/code>: Place in <code>models\/vae\/<\/code> and load with the Load VAE node.<\/li>\n<li><code>8-steps-dmd-lora.safetensors<\/code>: Place in <code>models\/loras\/<\/code> and connect with the LoraLoaderModelOnly node with a strength of 1.<\/li>\n<\/ul>\n<p>When using LoRA together, set the KSampler configuration to the <code>euler<\/code> sampler, <code>simple<\/code> scheduler, 8 steps, and CFG 1. In addition, <code>convert_to_comfy.py<\/code> for converting the proprietary Transformer folder into a single-file format and <code>quantize_convrot.py<\/code> for converting to W8A8 quantization format are provided within the repository.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-moss-image-sft-gguf\/\">MOSS-Image-SFT-GGUF 5.1B Text-to-Image Generation Model: 32GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/prunaai-qwen-image-2-1-remove-object\/\">PrunaAI Releases Object Removal LoRA for Qwen-Image-2.1<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/flux-3-image-klein-2\/\">Black Forest Labs Releases FLUX.3 Image Klein<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/28\/minimax-h3-character-swap-lora\/\">MiniMax-H3-Character-Swap-LoRA Character Swap LoRA Adapter: File List<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-openmoss-team-moss-image-en\/\">MOSS-Image family overview (2 articles)<\/a><\/li>\n<li><strong>Formats this model is available in<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">Safetensors format guide and models<\/a><\/li>\n<li><strong>Other models for the same task<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/models-by-task-en\/#task-image\">Other image generation models<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/lodestones\/Kromoss\">lodestones\/Kromoss (Hugging Face)<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/OpenMOSS-Team\/MOSS-Image-Base\">OpenMOSS-Team\/MOSS-Image-Base (Hugging Face)<\/a><\/li>\n<\/ul>\n<blockquote>\n<p><strong>This article contains unverified information.<\/strong> We will append an update note once it is confirmed by a primary source.<\/p>\n<\/blockquote>\n","protected":false},"excerpt":{"rendered":"<p>Learn about lodestones\/Kromoss, an open-weight text-to-image model fine-tuned from MOSS-Image-Base to reduce blurriness, including specs and usage.<\/p>\n","protected":false},"author":1,"featured_media":12616,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[203],"tags":[677,747,3788,3790,111,3720,1565,753],"class_list":["post-12617","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-image-video-and-audio","tag-comfyui-en","tag-diffusers-en","tag-lodestones-en","tag-lodestones-kromoss-en","tag-lora-en","tag-moss-image-en","tag-unverified","tag--en"],"lang":"en","translations":{"en":12617,"ja":12615},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12617","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=12617"}],"version-history":[{"count":3,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12617\/revisions"}],"predecessor-version":[{"id":17867,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12617\/revisions\/17867"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/12616"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=12617"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=12617"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=12617"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}