{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5","slug":"stable-diffusion-v1-5--stable-diffusion-v1-5","name":"stable-diffusion-v1-5","type":"model","url":"https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5","page_url":"https://unfragile.ai/stable-diffusion-v1-5--stable-diffusion-v1-5","categories":["image-generation"],"tags":["diffusers","safetensors","stable-diffusion","stable-diffusion-diffusers","text-to-image","arxiv:2207.12598","arxiv:2112.10752","arxiv:2103.00020","arxiv:2205.11487","arxiv:1910.09700","license:creativeml-openrail-m","endpoints_compatible","diffusers:StableDiffusionPipeline","region:us"],"pricing":{"model":"open_source","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_0","uri":"capability://image.visual.latent.space.text.to.image.generation.with.diffusion.sampling","name":"latent-space text-to-image generation with diffusion sampling","description":"Generates images from text prompts by iteratively denoising latent representations through a learned diffusion process. Uses a pre-trained CLIP text encoder to embed prompts into a shared semantic space, then conditions a UNet-based diffusion model operating in compressed latent space (via VAE) to progressively denoise Gaussian noise into coherent images over 20-50 sampling steps. Supports multiple schedulers (DDPM, PNDM, LMSDiscrete, EulerAncestralDiscrete) for speed/quality tradeoffs.","intents":["Generate photorealistic or artistic images from natural language descriptions","Create variations of images by adjusting sampling parameters and random seeds","Integrate image generation into applications without cloud API dependencies","Fine-tune or extend the model for domain-specific image synthesis tasks"],"best_for":["developers building offline-capable image generation features","researchers experimenting with diffusion model architectures","teams needing cost-effective, self-hosted image synthesis at scale","creators prototyping generative AI products without vendor lock-in"],"limitations":["Requires 4-8GB VRAM for inference; slower on CPU (30-120s per image vs 2-5s on GPU)","Latent space compression via VAE introduces subtle artifacts and loss of fine detail","Text understanding limited to CLIP's training data; struggles with complex spatial relationships or rare concepts","No built-in inpainting, outpainting, or image-to-image capabilities in base model (requires separate pipelines)","Deterministic only with fixed seed; no control over specific object placement or composition without additional guidance"],"requires":["Python 3.8+","PyTorch 1.9+ with CUDA 11.0+ (or CPU fallback, much slower)","diffusers library 0.10.0+","transformers library for CLIP text encoder","6GB+ free disk space for model weights (safetensors format ~4GB)","PIL/Pillow for image I/O"],"input_types":["text (prompt string, max ~77 tokens via CLIP tokenizer)","numeric (guidance_scale: 7.5-15 typical, num_inference_steps: 20-50, height/width: 512x512 or multiples of 64)"],"output_types":["PIL Image object","numpy array (uint8, shape [height, width, 3])","PNG/JPEG bytes"],"categories":["image-visual","generative-ai"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_1","uri":"capability://image.visual.classifier.free.guidance.with.prompt.weighting","name":"classifier-free guidance with prompt weighting","description":"Implements conditional image generation by blending unconditional and conditional noise predictions during diffusion sampling. At each denoising step, the model predicts noise for both the text prompt and an empty/null prompt, then interpolates between them using a guidance scale (typically 7.5-15) to amplify prompt adherence. This allows fine-grained control over image-prompt alignment without retraining, trading off diversity for fidelity.","intents":["Control the strength of prompt influence on generated images (weak guidance = more creative, strong = more literal)","Generate images that closely match specific text descriptions","Reduce unwanted artifacts by adjusting guidance scale per generation"],"best_for":["developers tuning image quality for specific use cases","users wanting control over creativity vs. prompt adherence tradeoff"],"limitations":["High guidance scales (>15) can produce oversaturated colors and unnatural textures","Guidance scale is a hyperparameter requiring manual tuning per prompt or domain","No per-token or per-phrase weighting; entire prompt weighted uniformly","Computational cost increases linearly with guidance scale (2x forward passes per step)"],"requires":["diffusers StableDiffusionPipeline with guidance_scale parameter support"],"input_types":["text (positive prompt)","float (guidance_scale, typical range 7.5-15)"],"output_types":["PIL Image"],"categories":["image-visual","generative-ai"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_10","uri":"capability://image.visual.memory.efficient.inference.with.attention.slicing.and.gradient.checkpointing","name":"memory-efficient inference with attention slicing and gradient checkpointing","description":"Reduces peak memory usage during inference by splitting attention computation across spatial dimensions (attention slicing) and enabling gradient checkpointing (recomputing activations instead of storing them). Attention slicing computes attention in chunks, reducing intermediate tensor sizes. Gradient checkpointing trades compute for memory by recomputing forward passes during backward passes (useful for fine-tuning). These optimizations are optional and can be enabled/disabled via pipeline configuration.","intents":["Generate images on GPUs with limited VRAM (e.g., RTX 3060 with 12GB)","Enable fine-tuning on consumer hardware by reducing memory footprint","Support higher resolutions or larger batch sizes on fixed hardware"],"best_for":["developers deploying on resource-constrained hardware","researchers fine-tuning models on consumer GPUs","applications requiring maximum memory efficiency"],"limitations":["Attention slicing reduces memory but increases latency by 10-20% due to chunking overhead","Gradient checkpointing is only useful for fine-tuning (not inference); adds 20-30% compute overhead","Memory savings are modest (typically 20-30%) and diminish with larger batch sizes","No automatic selection of optimal slicing strategy; requires manual tuning","Incompatible with some advanced features (e.g., xFormers optimization)"],"requires":["diffusers pipeline with enable_attention_slicing() method","PyTorch with gradient checkpointing support (optional)"],"input_types":["boolean (enable_attention_slicing, enable_gradient_checkpointing)"],"output_types":["PIL Image (same as without optimization)"],"categories":["image-visual","performance-optimization"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_11","uri":"capability://image.visual.xformers.integration.for.optimized.attention.computation","name":"xformers integration for optimized attention computation","description":"Integrates the xFormers library for memory-efficient and fast attention computation using fused kernels and approximations. xFormers provides optimized implementations of attention (FlashAttention, memory-efficient attention) that reduce memory usage by 30-50% and improve speed by 2-3x compared to standard PyTorch attention. Integration is automatic if xFormers is installed; no code changes required.","intents":["Accelerate image generation by 2-3x via optimized attention kernels","Reduce memory usage by 30-50% for larger batch sizes or resolutions","Enable real-time generation on consumer hardware"],"best_for":["developers optimizing inference latency for production","applications requiring real-time or interactive generation","teams deploying on specific GPU architectures (NVIDIA A100, RTX 40-series)"],"limitations":["xFormers requires CUDA and is not available on CPU or non-NVIDIA GPUs","xFormers support varies by GPU architecture; optimal performance on newer GPUs (A100, RTX 40-series)","xFormers is an external dependency with its own versioning and compatibility issues","Approximations in xFormers (e.g., FlashAttention) may introduce subtle quality differences","Installation can be complex due to CUDA version requirements"],"requires":["xFormers library (pip install xformers)","CUDA 11.0+ and compatible NVIDIA GPU","diffusers with xFormers support"],"input_types":["automatic (no user input required if xFormers is installed)"],"output_types":["PIL Image (same as without xFormers, but faster)"],"categories":["image-visual","performance-optimization"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_12","uri":"capability://code.generation.editing.lora.fine.tuning.support.for.efficient.model.adaptation","name":"lora fine-tuning support for efficient model adaptation","description":"Enables efficient fine-tuning via Low-Rank Adaptation (LoRA), which adds small trainable matrices to model weights without modifying the base model. LoRA reduces fine-tuning parameters by 100-1000x (e.g., 50M parameters instead of 860M for full fine-tuning), enabling training on consumer GPUs. LoRA weights are stored separately and can be merged into the base model or loaded dynamically during inference.","intents":["Fine-tune Stable Diffusion for domain-specific image generation (e.g., product photography, portraits)","Adapt the model to new styles or concepts with minimal training data and compute","Enable efficient multi-tenant systems where each user has a custom LoRA"],"best_for":["developers building customizable image generation systems","teams fine-tuning for specific domains or styles","researchers experimenting with model adaptation"],"limitations":["LoRA is less expressive than full fine-tuning; may not capture complex domain shifts","LoRA rank (typically 4-64) is a hyperparameter requiring tuning","Fine-tuning still requires a dataset of 100-1000 images; no true few-shot learning","LoRA weights are model-specific; cannot transfer between different base models","Merging LoRA into the base model is irreversible; requires storing both versions"],"requires":["diffusers with LoRA support (via peft library)","training dataset (100-1000 images)","training code (not provided in base model; requires external tools like Dreambooth or Kohya)"],"input_types":["image dataset (PNG/JPEG files)","text prompts (optional, for text-guided fine-tuning)"],"output_types":["LoRA weights (safetensors file, typically 1-10MB)"],"categories":["code-generation-editing","image-visual"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_2","uri":"capability://image.visual.multi.scheduler.diffusion.sampling.with.speed.quality.tradeoffs","name":"multi-scheduler diffusion sampling with speed-quality tradeoffs","description":"Provides pluggable noise schedulers (DDPM, PNDM, LMSDiscrete, EulerAncestralDiscrete, DPMSolverMultistep) that control the denoising trajectory and step count. Different schedulers trade off inference speed (fewer steps = faster) against image quality and diversity. DDPM is the original slow baseline; PNDM and Euler variants enable 20-30 step generation with minimal quality loss; DPMSolver achieves good results in 10-15 steps.","intents":["Generate images quickly for real-time or interactive applications (10-20 steps)","Maximize image quality when latency is not a constraint (50+ steps)","Experiment with different sampling strategies to find optimal speed-quality balance for a domain"],"best_for":["developers optimizing inference latency for production deployments","researchers benchmarking diffusion sampling strategies","applications requiring variable quality tiers (fast preview vs. high-quality final render)"],"limitations":["Fewer steps increases variance and may produce lower-quality or inconsistent results","Scheduler choice is not well-documented; requires empirical testing per use case","Some schedulers (e.g., DPMSolver) are newer and less battle-tested than DDPM","Scheduler switching requires reloading the pipeline; no dynamic switching mid-generation"],"requires":["diffusers library with scheduler implementations","knowledge of scheduler tradeoffs (not obvious from API)"],"input_types":["scheduler name (string: 'DDPM', 'PNDM', 'LMSDiscrete', 'EulerAncestralDiscrete', 'DPMSolverMultistep')","num_inference_steps (int, typically 20-50)"],"output_types":["PIL Image"],"categories":["image-visual","performance-optimization"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_3","uri":"capability://text.generation.language.clip.based.semantic.text.encoding.with.prompt.tokenization","name":"clip-based semantic text encoding with prompt tokenization","description":"Encodes text prompts into 768-dimensional embeddings using OpenAI's CLIP text encoder (ViT-L/14), which maps natural language to a shared semantic space with images. Tokenizes prompts using a BPE tokenizer with a 77-token context window, truncating or padding longer inputs. Embeddings are then used to condition the UNet diffusion model via cross-attention layers, enabling semantic understanding of arbitrary English prompts without task-specific training.","intents":["Convert natural language descriptions into semantic embeddings for image conditioning","Support flexible, open-vocabulary prompts without predefined class lists","Enable multi-modal understanding by leveraging CLIP's vision-language alignment"],"best_for":["developers building flexible text-to-image systems","applications requiring semantic understanding of user-provided descriptions"],"limitations":["77-token limit truncates long prompts; no built-in prompt expansion or summarization","CLIP tokenizer is BPE-based and may tokenize rare words or technical terms inefficiently","CLIP's training data (400M image-text pairs) introduces biases and limitations in understanding niche domains","No support for non-English languages (CLIP trained primarily on English)","Semantic understanding is coarse; CLIP struggles with spatial relationships, counts, and fine-grained attributes"],"requires":["transformers library with CLIP model","OpenAI CLIP model weights (~340MB)","text input in English"],"input_types":["text (English prompt string, max 77 tokens)"],"output_types":["torch.Tensor (shape [1, 77, 768] for batch size 1)"],"categories":["text-generation-language","image-visual"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_4","uri":"capability://image.visual.vae.based.latent.space.compression.and.reconstruction","name":"vae-based latent space compression and reconstruction","description":"Encodes images into a compressed latent space using a pre-trained Variational Autoencoder (VAE) with 4x4x4 spatial compression (512x512 image → 64x64x4 latent). The diffusion process operates in this latent space rather than pixel space, reducing memory requirements and computation by ~16x. After denoising, a VAE decoder reconstructs the latent back to pixel space. This two-stage approach (encode → diffuse → decode) is the core efficiency innovation enabling consumer-GPU inference.","intents":["Enable efficient image generation on consumer GPUs by reducing memory footprint","Support higher resolutions (512x512+) without requiring enterprise hardware","Compress image information for faster diffusion sampling"],"best_for":["developers deploying image generation on limited hardware (RTX 3060, RTX 4060)","applications requiring real-time or near-real-time generation","cost-sensitive deployments where GPU memory is a bottleneck"],"limitations":["VAE compression introduces subtle artifacts and loss of fine detail (especially thin lines, small text)","VAE decoder can produce blurry outputs if latent quality is poor","Compression is lossy; reconstructed images differ from originals (LPIPS ~0.1 typical)","VAE is fixed and not fine-tunable; cannot improve reconstruction quality without retraining","Latent space is not interpretable; cannot directly manipulate latents for semantic edits"],"requires":["pre-trained VAE model weights (~84MB)","diffusers AutoencoderKL implementation"],"input_types":["PIL Image or torch.Tensor (shape [batch, 3, 512, 512])"],"output_types":["torch.Tensor (latent, shape [batch, 4, 64, 64])","PIL Image (reconstructed, shape [batch, 3, 512, 512])"],"categories":["image-visual","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_5","uri":"capability://image.visual.negative.prompt.conditioning.for.artifact.suppression","name":"negative prompt conditioning for artifact suppression","description":"Allows specification of negative prompts (undesired attributes) that are subtracted from the guidance signal during diffusion sampling. Negative prompts are encoded via CLIP and their noise predictions are subtracted from the conditional predictions, effectively pushing the model away from undesired concepts. This is implemented as an extension of classifier-free guidance with separate guidance scales for positive and negative prompts.","intents":["Suppress unwanted visual artifacts or attributes (e.g., 'blurry, low quality')","Exclude specific objects or styles from generated images","Fine-tune image generation by specifying what NOT to generate"],"best_for":["developers refining image quality by eliminating common artifacts","users wanting more control over generated content"],"limitations":["Negative prompts are less effective than positive prompts; require careful wording","No principled way to weight negative vs. positive guidance; requires manual tuning","Negative prompts can conflict with positive prompts, leading to unpredictable results","Effectiveness varies widely depending on prompt specificity and model training"],"requires":["diffusers StableDiffusionPipeline with negative_prompt parameter"],"input_types":["text (negative prompt string, max 77 tokens)"],"output_types":["PIL Image"],"categories":["image-visual","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_6","uri":"capability://image.visual.deterministic.generation.with.seed.control","name":"deterministic generation with seed control","description":"Enables reproducible image generation by fixing random seeds for noise initialization and sampling. Setting a seed ensures the same image is generated for identical prompts and hyperparameters, critical for debugging, A/B testing, and user-facing features requiring consistency. Seeds are passed to PyTorch's random number generator and control both initial noise and stochastic sampling steps.","intents":["Generate reproducible images for testing and debugging","Enable A/B testing by comparing images with different prompts but same seed","Provide consistent results for user-facing applications"],"best_for":["developers building production image generation systems","researchers benchmarking and comparing models","applications requiring audit trails or reproducibility"],"limitations":["Seed reproducibility is only guaranteed within the same hardware, PyTorch version, and scheduler","Different GPUs or CPU implementations may produce slightly different results due to floating-point non-determinism","Seed does not control CLIP text encoding randomness (if any); only diffusion sampling","No built-in seed management for batch generation; requires manual seed incrementing"],"requires":["torch.manual_seed() or diffusers generator parameter","consistent hardware and software versions"],"input_types":["int (seed value, typically 0-2^32-1)"],"output_types":["PIL Image (deterministic given seed)"],"categories":["image-visual","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_7","uri":"capability://image.visual.batch.image.generation.with.memory.efficient.processing","name":"batch image generation with memory-efficient processing","description":"Supports generating multiple images in parallel by batching prompts and noise tensors, reducing per-image overhead and improving GPU utilization. Batch size is limited by available VRAM; typical batch sizes are 1-4 on consumer GPUs (8GB VRAM) and 8-16 on high-end GPUs (24GB+). Batching is implemented via standard PyTorch tensor operations with no special optimization; memory usage scales linearly with batch size.","intents":["Generate multiple images efficiently for applications requiring bulk image synthesis","Improve GPU utilization by amortizing fixed overhead across multiple images","Enable parallel exploration of prompt variations"],"best_for":["applications generating images in bulk (e.g., dataset creation, content generation)","developers optimizing inference throughput on fixed hardware"],"limitations":["Memory usage scales linearly with batch size; no built-in memory optimization","Batch size is limited by VRAM; no automatic batching or spilling to CPU","All prompts in a batch must be the same length (padded to max length), wasting tokens","No built-in progress tracking or cancellation for long batch jobs","Batching adds minimal speedup per image (typically 10-20% improvement) due to fixed overhead"],"requires":["sufficient VRAM for batch size (roughly 1.5GB per image at 512x512)"],"input_types":["list of text prompts (all same length or padded)","int (batch_size, limited by VRAM)"],"output_types":["list of PIL Images"],"categories":["image-visual","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_8","uri":"capability://safety.moderation.safetensors.format.model.loading.with.security.validation","name":"safetensors format model loading with security validation","description":"Loads model weights from the safetensors format, a safer alternative to pickle that prevents arbitrary code execution during deserialization. Safetensors is a simple binary format with explicit type information, enabling validation of tensor shapes and dtypes before loading. The diffusers library automatically detects and loads safetensors files, falling back to PyTorch .bin format if unavailable.","intents":["Load model weights safely without risk of code injection","Validate model integrity before loading","Enable faster model loading via memory-mapped safetensors files"],"best_for":["developers prioritizing security in model loading","applications loading untrusted or third-party model weights","teams with strict security requirements"],"limitations":["Safetensors support is optional; not all models provide safetensors versions","Fallback to .bin (pickle) format is automatic, potentially loading unsafe models silently","No built-in signature verification or checksum validation","Safetensors format is newer and less widely supported than pickle in some tools"],"requires":["safetensors library (optional but recommended)","diffusers library with safetensors support"],"input_types":["file path (string, .safetensors or .bin format)"],"output_types":["loaded model weights (torch.nn.Module)"],"categories":["safety-moderation","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__cap_9","uri":"capability://image.visual.cross.attention.visualization.and.prompt.token.attribution","name":"cross-attention visualization and prompt token attribution","description":"Provides access to cross-attention maps (attention weights between text tokens and image spatial locations) during diffusion sampling, enabling visualization of which image regions correspond to which prompt tokens. Cross-attention maps are computed at each diffusion step and can be extracted via hooks or custom pipeline modifications. This enables interpretability and debugging of prompt-image alignment.","intents":["Visualize which image regions correspond to which prompt tokens","Debug prompt understanding and identify misalignments","Implement prompt-guided editing by manipulating attention maps"],"best_for":["researchers studying diffusion model interpretability","developers debugging prompt-image misalignment issues","advanced users implementing custom editing or control mechanisms"],"limitations":["Cross-attention extraction requires custom pipeline code or hooks; not exposed in standard API","Attention maps are high-dimensional (num_tokens × height × width × num_heads); visualization requires dimensionality reduction","Attention maps change at each diffusion step; no single 'final' attribution","Attention maps are not causal; they don't directly explain which tokens influenced which pixels","Extracting attention maps adds computational overhead and requires model introspection"],"requires":["custom diffusers pipeline code with attention hooks","understanding of cross-attention mechanism and transformer architecture"],"input_types":["diffusion pipeline with attention hooks installed"],"output_types":["torch.Tensor (attention maps, shape [num_tokens, height, width, num_heads])","visualization (heatmap or overlay image)"],"categories":["image-visual","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hf-model-stable-diffusion-v1-5--stable-diffusion-v1-5__headline","uri":"capability://image.visual.text.to.image.generation.model","name":"text-to-image generation model","description":"Stable Diffusion v1.5 is a powerful text-to-image generation model that allows users to create high-quality images from textual descriptions, making it ideal for artists, designers, and content creators looking to visualize concepts quickly.","intents":["best text-to-image model","text-to-image generation for creative projects","top AI models for generating images from text","how to create images from text descriptions","best tools for AI art generation"],"best_for":["artists","designers","content creators"],"limitations":["may require fine-tuning for specific styles"],"requires":["text input"],"input_types":["text"],"output_types":["images"],"categories":["image-visual"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":54,"verified":false,"data_access_risk":"low","permissions":["Python 3.8+","PyTorch 1.9+ with CUDA 11.0+ (or CPU fallback, much slower)","diffusers library 0.10.0+","transformers library for CLIP text encoder","6GB+ free disk space for model weights (safetensors format ~4GB)","PIL/Pillow for image I/O","diffusers StableDiffusionPipeline with guidance_scale parameter support","diffusers pipeline with enable_attention_slicing() method","PyTorch with gradient checkpointing support (optional)","xFormers library (pip install xformers)"],"failure_modes":["Requires 4-8GB VRAM for inference; slower on CPU (30-120s per image vs 2-5s on GPU)","Latent space compression via VAE introduces subtle artifacts and loss of fine detail","Text understanding limited to CLIP's training data; struggles with complex spatial relationships or rare concepts","No built-in inpainting, outpainting, or image-to-image capabilities in base model (requires separate pipelines)","Deterministic only with fixed seed; no control over specific object placement or composition without additional guidance","High guidance scales (>15) can produce oversaturated colors and unnatural textures","Guidance scale is a hyperparameter requiring manual tuning per prompt or domain","No per-token or per-phrase weighting; entire prompt weighted uniformly","Computational cost increases linearly with guidance scale (2x forward passes per step)","Attention slicing reduces memory but increases latency by 10-20% due to chunking overhead","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7842535013240095,"quality":0.5,"ecosystem":0.5000000000000001,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.35,"quality":0.2,"ecosystem":0.1,"match_graph":0.3,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:22.766Z","last_scraped_at":"2026-05-03T14:22:49.651Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":1481468,"model_likes":1091}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=stable-diffusion-v1-5--stable-diffusion-v1-5","compare_url":"https://unfragile.ai/compare?artifact=stable-diffusion-v1-5--stable-diffusion-v1-5"}},"signature":"9VbAxCGShlri5orFdxsU5oQZMqRCWmh3ugFme9BmnC7RQ7jq8JAGYQGdWDI82ZInMeisfW3+/wV1DMQAAhplDQ==","signedAt":"2026-06-21T09:00:31.821Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/stable-diffusion-v1-5--stable-diffusion-v1-5","artifact":"https://unfragile.ai/stable-diffusion-v1-5--stable-diffusion-v1-5","verify":"https://unfragile.ai/api/v1/verify?slug=stable-diffusion-v1-5--stable-diffusion-v1-5","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}