{"id":36058,"date":"2026-08-28T05:03:42","date_gmt":"2026-08-28T05:03:42","guid":{"rendered":"https:\/\/www.sozee.ai\/resources\/fooocus-vs-automatic1111-consistent-images\/"},"modified":"2026-09-02T13:17:05","modified_gmt":"2026-09-02T13:17:05","slug":"fooocus-vs-automatic1111-consistent-images","status":"publish","type":"post","link":"https:\/\/www.sozee.ai\/resources\/fooocus-vs-automatic1111-consistent-images\/","title":{"rendered":"Fooocus vs Automatic1111 vs Sozee: Face Consistency"},"content":{"rendered":"<h2 id=\"key-takeaways\">Key Takeaways<\/h2>\n<ul>\n<li>Fooocus and Automatic1111 both introduce face-lock drift that breaks consistency across 50\u2013200 images per month.<\/li>\n<li>Neither tool offers reliable batch reproducibility or built-in scheduling for daily posting workflows.<\/li>\n<li>LoRA training, prompt expansion, and extension conflicts create hidden engineering costs that grow with production volume.<\/li>\n<li>Brand deals, SFW-to-NSFW arcs, and virtual influencer campaigns all need locked identity that these tools cannot guarantee.<\/li>\n<li>Creators who want to eliminate drift and re-shoots can <a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\">start a free trial with Sozee<\/a> and lock likeness from the first frame.<\/li>\n<\/ul>\n<h2>Face Consistency Showdown: Fooocus vs Automatic1111 vs Sozee<\/h2>\n<p>Fooocus and Automatic1111 were not built to guarantee the same face across a commercial image series. Each tool\u2019s identity controls introduce drift at different points in the pipeline.<\/p>\n<p>A custom LoRA trained on 15\u201330 varied images <a href=\"https:\/\/dev.to\/miriam_alonso_01\/how-ai-headshot-generators-actually-work-lora-15-selfies-and-the-overfitting-problem-43p\" target=\"_blank\" rel=\"noindex nofollow\">often overfits, reproducing training details instead of generalizing reliably across many generations<\/a>. This method delivers high-fidelity serial character output, but it demands cloud-GPU time and careful dataset preparation. A single selfie reference overfits immediately, and that overhead repeats every time a new character or major style change is needed.<\/p>\n<p>Once a LoRA is trained, stacking it with ControlNet introduces a second failure point. <a href=\"https:\/\/iimagined.ai\/blog\/controlnet-complete-guide-2026\" target=\"_blank\" rel=\"noindex nofollow\">ControlNet manages per-generation composition controls while LoRA manages recurring identity<\/a>. <a href=\"https:\/\/hackernoon.com\/flux2-klein-9b-matchingpose-a-guide-to-identity-preserving-pose-transfer\" target=\"_blank\" rel=\"noindex nofollow\">LoRA strength of 0.9\u20131.1 is recommended during inference to balance identity preservation with pose control<\/a>. That balance requires manual calibration that shifts with every new scene or outfit.<\/p>\n<p>Fooocus adds a structural problem on top of this. Fooocus processes user prompts so the input prompt and the actual prompt sent to the model can differ. For Pony or Illustrious models, prompt expansion can introduce tokens that push generated faces and styles away from earlier outputs. <a href=\"https:\/\/offlinecreator.com\/compare\/fooocus-vs-forge\" target=\"_blank\" rel=\"noindex nofollow\">Fooocus also hides sampling steps, CFG scale, and scheduler by default<\/a>, which removes the controls needed to hold a character\u2019s likeness steady at scale.<\/p>\n<p>Automatic1111 exposes those controls but creates its own drift surface. <a href=\"https:\/\/zview.space\/articles\/comfyui-workflow-vs-automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">Output shifts in Automatic1111 originate from sampler changes, scheduler changes, VAE differences, latent upscale order, ControlNet weight drift, denoise strength variations, extension interactions, LoRA stacking order, and prompt edits, even when users believe the workflow is unchanged<\/a>. <a href=\"https:\/\/howtosd.com\/how-to-use-controlnet-in-automatic1111-part-4-reference-preprocessor-to-create-slight-variations-in-images\/\" target=\"_blank\" rel=\"noindex nofollow\">Stacking ControlNet, LoRA, and img2img adjustments in Automatic1111 is used to reduce face and composition variation across runs and produce more consistent outputs<\/a>, but each layer adds more variables to manage.<\/p>\n<p>The table below shows how each tool\u2019s approach to face-locking creates different failure points and why neither can deliver commercial-scale consistency.<\/p>\n<table>\n<thead>\n<tr>\n<th>Criterion<\/th>\n<th>Fooocus<\/th>\n<th>Automatic1111<\/th>\n<th>Sozee<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Face-lock method<\/td>\n<td><a href=\"https:\/\/offlinecreator.com\/compare\/fooocus-vs-automatic1111-vs-comfyui-2026\" target=\"_blank\" rel=\"noindex nofollow\">Preset-driven LoRA; good reproducibility for simple use<\/a><\/td>\n<td><a href=\"https:\/\/offlinecreator.com\/compare\/fooocus-vs-automatic1111-vs-comfyui-2026\" target=\"_blank\" rel=\"noindex nofollow\">Medium reproducibility; requires user discipline<\/a><\/td>\n<td>Direction-based likeness lock; same face every frame by design<\/td>\n<\/tr>\n<tr>\n<td>Prompt drift risk<\/td>\n<td><a href=\"https:\/\/offlinecreator.com\/compare\/fooocus-vs-forge\" target=\"_blank\" rel=\"noindex nofollow\">High, because prompt processing can alter the actual prompt sent to the model<\/a><\/td>\n<td><a href=\"https:\/\/zview.space\/articles\/comfyui-workflow-vs-automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">High, because prompt sensitivity shifts geometry and facial identity<\/a><\/td>\n<td>None, because likeness is a locked asset, not a prompt variable<\/td>\n<\/tr>\n<tr>\n<td>Parameter visibility<\/td>\n<td><a href=\"https:\/\/offlinecreator.com\/compare\/fooocus-vs-forge\" target=\"_blank\" rel=\"noindex nofollow\">Hidden by default, with CFG, steps, and scheduler not exposed<\/a><\/td>\n<td><a href=\"https:\/\/zview.space\/articles\/comfyui-workflow-vs-automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">Fully exposed but creates a multi-variable drift surface<\/a><\/td>\n<td>Five directed dimensions replace parameter tuning<\/td>\n<\/tr>\n<tr>\n<td>LoRA training required<\/td>\n<td>Yes, with dataset preparation and compute resources per character<\/td>\n<td>Yes, with dataset preparation and compute resources per character<\/td>\n<td>No, with three photos or zero photos using character generation<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><strong>See how Sozee\u2019s direction-based likeness lock removes the drift shown in this comparison and start your free trial.<\/strong><\/a><\/p>\n<figure style=\"text-align: center;\"><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><img src=\"https:\/\/cdn.aigrowthmarketer.co\/1762997925636-7453a7a8b2ad.png\" alt=\"Sozee AI Platform\" style=\"max-height: 500px;\" loading=\"lazy\" decoding=\"async\"><\/a><figcaption><em>Sozee AI Platform<\/em><\/figcaption><\/figure>\n<h2>Batch Speed and Daily Posting: Repeatability Beats Raw FPS<\/h2>\n<p>Raw generation speed on identical hardware stays close between Fooocus and Automatic1111. <a href=\"https:\/\/lilys.ai\/en\/notes\/comfyui-20260115\/sdxl-speed-test-comfyui-invokeai-a1111\" target=\"_blank\" rel=\"noindex nofollow\">On an RTX 4090 running SDXL at 1024\u00d71024, Automatic1111 completes a single txt2img in about 36 seconds<\/a>. A sequential batch of images takes similar time in both tools on the same hardware.<\/p>\n<p>The real gap appears in repeatability, not speed. <a href=\"https:\/\/yuzec.com\/blog\/comfyui-vs-a1111-vs-fooocus\" target=\"_blank\" rel=\"noindex nofollow\">Automatic1111\u2019s pipeline architecture can require full re-runs even when only the seed changes<\/a>, which reduces efficiency for near-variant batches. <a href=\"https:\/\/yuzec.com\/blog\/comfyui-vs-a1111-vs-fooocus\" target=\"_blank\" rel=\"noindex nofollow\">Fooocus provides no workflow portability or equivalent to saved generation parameters<\/a>, so exact pipeline reproduction across sessions becomes unreliable.<\/p>\n<p><a href=\"https:\/\/offlinecreator.com\/compare\/comfyui-vs-automatic1111-vs-invokeai-vs-fooocus-2026\" target=\"_blank\" rel=\"noindex nofollow\">Latency in local Stable Diffusion UIs can vary significantly on the same GPU when sampler, steps, resolution, and attention backends change<\/a>. For a creator producing 50\u2013200 images per month across multiple outfits, settings, and expressions, that variance turns into broken deliverables instead of a minor benchmark detail.<\/p>\n<p>Fooocus\u2019s prompt expansion introduces another batch hazard. <a href=\"https:\/\/localaimaster.com\/blog\/fooocus-guide\" target=\"_blank\" rel=\"noindex nofollow\">LoRA effects produce no visible change unless the exact trigger word appears in the prompt<\/a>, and <a href=\"https:\/\/localaimaster.com\/blog\/fooocus-guide\" target=\"_blank\" rel=\"noindex nofollow\">style presets fail to apply when an incompatible checkpoint is loaded<\/a>. A batch of ten images can drift silently in the middle of a run if any of these conditions fail on even one generation.<\/p>\n<h2>Revenue Workflows: Where Drift Turns Into Lost Income<\/h2>\n<p>The consistency failures above map directly to lost revenue in three common production scenarios.<\/p>\n<p><strong>Brand deal deliverables.<\/strong> A sponsorship brief usually requires a product in multiple settings, outfits, and angles that all look like the same person on the same day. <a href=\"https:\/\/aiphotogenerator.net\/blog\/2026\/05\/control-net-ai\" target=\"_blank\" rel=\"noindex nofollow\">OpenPose solves posture control but not identity control<\/a>. <a href=\"https:\/\/aiphotogenerator.net\/blog\/2026\/05\/control-net-ai\" target=\"_blank\" rel=\"noindex nofollow\">Users still need prompt discipline, reference-based methods, or another control layer for full identity consistency<\/a>. Assembling that stack for a paid campaign in Automatic1111 requires <a href=\"https:\/\/freevisualtools.com\/blog\/consistent-character-ai-images\" target=\"_blank\" rel=\"noindex nofollow\">an evening of setup for users who have not already installed the ControlNet extension and preprocessor models<\/a>. Fooocus offers no comparable depth. Sozee\u2019s Photo Shoot feature takes one image and builds a locked, coherent set of up to ten around it. Identity, outfit, and environment stay constant while angle, pose, and expression change. The sponsor\u2019s product drops into the Object slot, and the brief is delivered in an afternoon.<\/p>\n<figure style=\"text-align: center;\"><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><img src=\"https:\/\/sozee.ai\/wp-content\/uploads\/2025\/11\/Sozee-60-Seconds-To-Generate-Content-White.gif\" alt=\"GIF of Sozee Platform Generating Images Based On Inputs From Creator on a White Background\" style=\"max-height: 500px;\" loading=\"lazy\" decoding=\"async\"><\/a><figcaption><em>GIF of Sozee Platform Generating Images Based On Inputs From Creator on a White Background<\/em><\/figcaption><\/figure>\n<p><strong>SFW-to-NSFW arcs.<\/strong> Neither Fooocus nor Automatic1111 provides a structured ramp between content tiers. Fooocus is <a href=\"https:\/\/localaimaster.com\/blog\/fooocus-guide\" target=\"_blank\" rel=\"noindex nofollow\">in maintenance mode with no plans for new architectures<\/a>. Automatic1111\u2019s <a href=\"https:\/\/vantaige.io\/ai-tool\/automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">tab-based UI makes it difficult to construct workflows that branch or apply conditional logic<\/a>. Sozee\u2019s Photo Shoot builds a full SFW-to-NSFW arc with pacing and ceiling set by the creator, with no re-engineering between tiers.<\/p>\n<p><strong>Virtual influencer consistency.<\/strong> <a href=\"https:\/\/local-llm.net\/compare\/comfyui-vs-automatic1111-vs-forge\" target=\"_blank\" rel=\"noindex nofollow\">Automatic1111\u2019s extension ecosystem creates dependency conflicts that are common, which makes large extension stacks brittle for production workflows that require locked likeness across hundreds of images<\/a>. <a href=\"https:\/\/localaimaster.com\/blog\/fooocus-guide\" target=\"_blank\" rel=\"noindex nofollow\">Fooocus does not support Roop for advanced face swapping or custom training scripts that virtual creators rely on to maintain locked likeness across hundreds of images<\/a>. Sozee generates an original character, locks her likeness, builds her world once, and schedules her to post daily in a single platform.<\/p>\n<p><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><strong>Build your virtual influencer with the locked likeness and daily posting workflow described here and start your free trial.<\/strong><\/a><\/p>\n<h2>Hidden Costs: Setup Time, Re-shoots, and Scale Limits<\/h2>\n<p>The visible costs of Fooocus and Automatic1111 stay low, while the hidden costs do not.<\/p>\n<p><a href=\"https:\/\/offlinecreator.com\/tool\/automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">Automatic1111\u2019s last major release was v1.10.1 in February 2025, after which active development slowed significantly<\/a>. <a href=\"https:\/\/vantaige.io\/ai-tool\/automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">Windows users regularly encounter CUDA version mismatches, Python path conflicts, and virtual environment errors during installation<\/a>. <a href=\"https:\/\/offlinecreator.com\/tool\/automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">Automatic1111 requires a tested Python baseline of 3.10.6 for stable extension operation, and moving to Python 3.11 introduces compatibility risks<\/a>. <a href=\"https:\/\/vantaige.io\/ai-tool\/automatic1111\" target=\"_blank\" rel=\"noindex nofollow\">The version transition produced extension compatibility issues, and numerous open issues and pull requests have accumulated with limited merge activity<\/a>.<\/p>\n<p>Fooocus cuts setup time but trades it for capability loss. <a href=\"https:\/\/onegen.ai\/project\/fooocus-guide-high-quality-ai-image-generation-made-simple\" target=\"_blank\" rel=\"noindex nofollow\">Fooocus supports only custom checkpoints and LoRAs rather than dozens of extensions or scripted workflows for advanced customization<\/a>. The maintenance-mode status mentioned earlier means Fooocus will never support newer architectures like Flux or SD 3.5. <a href=\"https:\/\/localaimaster.com\/blog\/fooocus-guide\" target=\"_blank\" rel=\"noindex nofollow\">Even the actively maintained mashb1t 1-Up Edition fork remains limited to SDXL models as of August 2026<\/a>.<\/p>\n<p>The re-shoot cost hits revenue most directly. When face drift breaks a deliverable, the creator either regenerates the full set and pays again in GPU time and hours or delivers inconsistent assets and risks the brand relationship. Sozee removes re-shoots by design. Every setting, outfit, and object is saved as a reusable asset. A room built from up to four reference photos remains the same room across every future shoot, so every shoot makes the next one faster.<\/p>\n<h2>Decision Point: When Creators Outgrow Fooocus and Automatic1111<\/h2>\n<p>Fooocus works for exploratory image generation where consistency does not matter. Automatic1111 suits technically skilled users who need maximum extension flexibility and accept maintenance overhead. Both tools hit the same ceiling when a creator needs the same face, reliably, across 50 or more commercial images per month.<\/p>\n<p>The routing criteria for switching to Sozee fall into three groups. First come consistency failures, such as face drift breaking brand deal deliverables or daily posting schedules, extension conflicts in Automatic1111 causing output shifts between sessions, or Fooocus prompt expansion silently altering character identity mid-batch. Second comes workflow overhead, where LoRA training time and dataset preparation consume hours that should go to content creation. Third comes missing infrastructure, such as a required SFW-to-NSFW arc with a structured ramp or the need for scheduling, analytics, and multi-platform publishing in the same workflow as generation. Any one of these groups signals that the creator has outgrown both tools\u2019 architectural limits.<\/p>\n<p>Sozee replaces both tools with a direction-based studio. Creators control five dimensions per frame, use Photo Shoot for locked sets, rely on a native scheduler connected to Instagram, TikTok, X, Facebook, Reddit, and Fanvue, and track analytics that separate Sozee-posted performance from manually posted content.<\/p>\n<figure style=\"text-align: center;\"><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><img src=\"https:\/\/cdn.aigrowthmarketer.co\/1762997859947-4a2e298c7c02.png\" alt=\"Creator Onboarding For Sozee AI\" style=\"max-height: 500px;\" loading=\"lazy\" decoding=\"async\"><\/a><figcaption><em>Creator Onboarding<\/em><\/figcaption><\/figure>\n<p><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><strong>If these criteria match your workflow, start your free trial and remove the drift that is costing you deliverables.<\/strong><\/a><\/p>\n<h2>Frequently Asked Questions<\/h2>\n<h3>How do ControlNet and LoRA methods in Automatic1111 and Fooocus still produce identity drift at scale?<\/h3>\n<p>LoRA encodes a character\u2019s identity into model weights, and its strength must be balanced against ControlNet\u2019s pose or composition signals at inference time. As discussed earlier, LoRA strength must be manually rebalanced for each new scene or outfit. This calibration problem compounds when combined with Automatic1111\u2019s other drift sources, including sampler changes, scheduler changes, VAE differences, ControlNet weight settings, and LoRA stacking order, any of which can shift facial identity even with a fixed seed. In Fooocus, prompt processing can alter the tokens sent to the model, which can change face, style, and body characteristics. At 50 or more images per month, these variables combine into visible inconsistency across a content series.<\/p>\n<h3>What failure rates appear when scaling these tools to 50\u2013200 images per month?<\/h3>\n<p>Neither tool publishes a formal drift rate, yet the documented failure mechanisms make consistent output at that volume structurally unlikely. Automatic1111\u2019s process memory lives partly in UI state, extension settings, and operator recall, which makes exact reproduction of multi-step workflows difficult after even two days. As noted in the batch speed analysis, Fooocus\u2019s lack of workflow portability means pipelines cannot be reliably reproduced across sessions. This limitation becomes critical when a creator must match yesterday\u2019s output for a multi-day campaign. Latency on the same GPU can vary significantly when sampler, steps, resolution, or attention backends change, so a batch that looked consistent on day one may not reproduce on day three. For a creator posting daily, that variance turns into re-generation time, missed deadlines, or inconsistent deliverables.<\/p>\n<h3>Can face-swap or reference adapters in either tool guarantee brand-consistent output for paid campaigns?<\/h3>\n<p>No. Face-swap methods such as Roop introduce their own error surface. A 2026 study on face-swapping models found that Ghost-v2 could incorrectly replace facial attributes on partially occluded subjects, which shows that face-swap pipelines introduce identity or attribute errors in edge cases. Fooocus does not support Roop natively. In Automatic1111, Roop and similar extensions are community-maintained and face the same dependency conflict and breakage issues that affect the broader extension ecosystem. IP-Adapter preserves identity more strongly than Reference-only ControlNet but still requires combination with a trained LoRA for production-grade consistency, which adds back the training overhead. OpenPose, the most common pose-control method, solves posture but not identity, so users still need additional control layers for full facial consistency across a paid campaign deliverable.<\/p>\n<h3>How does Sozee remove extension and training overhead while supporting SFW-to-NSFW arcs?<\/h3>\n<p>Sozee replaces the LoRA training pipeline with a three-photo cast process. Creators upload three photos and Sozee reconstructs the likeness instantly, with no training, no dataset preparation, and no GPU cost. For creators who want a fully original character, the AI Character Builder generates a face that has never existed and locks it from the first frame, with no source photos required. Identity becomes a locked asset that persists across every shoot, setting, and output tier instead of a prompt variable or model weight. The SFW-to-NSFW arc runs through Photo Shoot, which builds a coherent set of up to ten images from a single frame with the content ramp and ceiling set by the creator. Every element, including setting, outfit, and object, is saved and reusable, so the arc can be extended or repeated without rebuilding the workflow.<\/p>\n<h2>Conclusion<\/h2>\n<p>Fooocus and Automatic1111 both leave creators exposed to drift. Fooocus hides the controls needed to lock identity and silently rewrites prompts. Automatic1111 exposes every variable and turns all of them into potential sources of inconsistency. Neither tool includes scheduling, analytics, or a structured content ramp, and both demand engineering overhead that grows with the creator\u2019s ambition instead of shrinking as the workflow matures.<\/p>\n<p>Sozee is the only studio that locks likeness for daily posting and brand deals without training, complex setup, or re-shoots. Cast a character in minutes. Direct five dimensions per frame. Build a month of content from one image. Schedule it across every platform from the same interface so every shoot makes the next one faster.<\/p>\n<p><a href=\"https:\/\/app.sozee.ai\/sign-up\" target=\"_blank\"><strong>Get started and stop losing brand deals to face drift while you build a consistent creator identity.<\/strong><\/a><\/p>\n<section data-read-next=\"true\">\n<h2>Read Next<\/h2>\n<ul>\n<li><a href=\"https:\/\/sozee.ai\/resources\/fooocus-alternatives-consistent-ai-characters\" target=\"_blank\">Fooocus Alternatives for Consistent AI Character Images 2026<\/a><\/li>\n<li><a href=\"https:\/\/sozee.ai\/resources\/fooocus-alternatives-consistent-characters\" target=\"_blank\">Fooocus Alternatives for Consistent Character Generation<\/a><\/li>\n<li><a href=\"https:\/\/sozee.ai\/resources\/fooocus-alternatives-daily-influencer\" target=\"_blank\">Best Fooocus Alternatives for Daily Influencer Posting<\/a><\/li>\n<li><a href=\"https:\/\/sozee.ai\/resources\/fooocus-alternatives-photorealistic-influencer-tools\" target=\"_blank\">Fooocus Alternatives for Photorealistic Influencer Tools<\/a><\/li>\n<li><a href=\"https:\/\/sozee.ai\/resources\/best-ai-face-generation-tools\" target=\"_blank\">Best AI Face Generation Tools Compared to Alternatives<\/a><\/li>\n<\/ul>\n<\/section>\n","protected":false},"excerpt":{"rendered":"<p>Tired of identity drift at scale? See how Sozee beats Fooocus &#038; Automatic1111 for consistent, revenue-ready AI content. Try Sozee today.<\/p>\n","protected":false},"author":2,"featured_media":36057,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2,8,5],"tags":[36],"class_list":["post-36058","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-photos","category-automation","category-tools","tag-character-consistency"],"_links":{"self":[{"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/posts\/36058","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/comments?post=36058"}],"version-history":[{"count":1,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/posts\/36058\/revisions"}],"predecessor-version":[{"id":42548,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/posts\/36058\/revisions\/42548"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/media\/36057"}],"wp:attachment":[{"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/media?parent=36058"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/categories?post=36058"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sozee.ai\/resources\/wp-json\/wp\/v2\/tags?post=36058"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}