GITNUXSOFTWARE ADVICE
Top 10 Best AI Ebony Black Skin Female Generator of 2026
An editorial ranking of ai ebony black skin female generator tools compares accuracy, controls, and tradeoffs for creators and visual teams.
How we ranked these tools
Core product claims cross-referenced against official documentation, changelogs, and independent technical reviews.
Analyzed video reviews and hundreds of written evaluations to capture real-world user experiences with each tool.
AI persona simulations modeled how different user types would experience each tool across common use cases and workflows.
Final rankings reviewed and approved by our editorial team with authority to override AI-generated scores based on domain expertise.
Score: Features 40% · Ease 30% · Value 30%
Gitnux may earn a commission through links on this page — this does not influence rankings. Editorial policy
Editor’s top 3 picks
Three quick recommendations before you dive into the full comparison below — each one leads on a different dimension.
RAWSHOT AI
RAWSHOT AI replaces the category's empty text box with a seven-step visual configuration: product, model, supporting garments, styling, background, lighting, and composition. Saved Stacks preserve those selections for repeatable catalogue treatments, while each setting remains visible and editable.
Built for indie labels, DTC apparel operators, marketplace sellers, and enterprise fashion teams needing consistent synthetic on-model imagery for collections, repeat listings, or large catalogue runs..
Midjourney
Editor pickStyle References and Moodboards preserve campaign-specific visual direction across repeated Black female portrait generations.
Built for fits when creative teams need polished Black female portrait concepts with detailed art direction and manual review..
Stable Diffusion
Editor pickDownloadable model weights enable local portrait pipelines with custom checkpoints and application-specific preprocessing.
Built for fits when teams need controllable portraits of Black women across local workflows and API calls..
Comparison Table
RAWSHOT AI
Block-based AI fashion photographyRAWSHOT AI creates original on-model fashion images and short videos by combining selectable synthetic models, garments, lighting, backgrounds, poses, and camera views.
RAWSHOT AI replaces the category's empty text box with a seven-step visual configuration: product, model, supporting garments, styling, background, lighting, and composition. Saved Stacks preserve those selections for repeatable catalogue treatments, while each setting remains visible and editable.
RAWSHOT AI offers more than 1,800 licence-free synthetic models, including over 1,200 adults and more than 600 children; no child was cast, photographed, or used as a likeness reference. A private model builder exposes ten attributes for women and eleven for men, while users can combine up to four garments, choose from 15 image frames, and produce stills at 2K or 4K. Every output includes C2PA credentials, visible and cryptographic watermarking, AI-labelled metadata, and a per-image attribute record.
The main tradeoff is a deliberately bounded workflow: users cannot enter free text, and RAWSHOT AI ships with one garment-accurate image style rather than a broad visual treatment library. That structure works well for a DTC label preparing consistent on-model images for 10–200 SKUs, while teams seeking experimental portraits, a specific real person, or non-fashion imagery will find the product limiting.
- +Full commercial rights forever, with no recurring licensing on library models.
- +More than 1,800 licence-free synthetic models, including over 600 children's models; no child was cast, photographed, or used as a likeness reference.
- +Saved Stacks support repeatable catalogue treatments, while the REST API matches the browser interface from single images to 10,000-plus runs.
- +C2PA credentials, layered watermarking, AI labels, and per-image audit trails support documented content handling.
- –No free-text input means users cannot improvise beyond the available selectable blocks.
- –The product ships with one accurate image style, so stylised or graded treatments require post-production.
- –Video is limited to three five-second scenes at 720p or 1080p.
- –RAWSHOT AI is built for apparel, footwear, and accessories rather than general-purpose image generation.
DTC apparel brands
Create consistent launch imagery across collections
Consistent collection imagery
Marketplace sellers
Generate on-model product listings
More complete product listings
Show 2 more scenarios
Kidswear labels
Show childrenswear on synthetic models
Lower-risk kidswear imagery
Brands select from over 600 children's synthetic models without casting, photographing, or using a child's likeness reference.
Fashion platform teams
Run catalogue production through API
Scalable image operations
Platform teams import products and trigger browser-equivalent generation workflows across thousands of catalogue items.
Best for: Indie labels, DTC apparel operators, marketplace sellers, and enterprise fashion teams needing consistent synthetic on-model imagery for collections, repeat listings, or large catalogue runs.
Midjourney
specialistAI image generator accessed through Discord and web interface supporting ethnic and skin-tone specific text prompts.
Style References and Moodboards preserve campaign-specific visual direction across repeated Black female portrait generations.
Creative directors can guide complexion, hair texture, wardrobe, lighting, framing, and expression through structured prompts and reference images. Midjourney supports image blending, region changes, zooming, panning, upscaling, and variations, which helps refine portrait composition without restarting each generation. Style References preserve a selected visual language across multiple outputs, while Moodboards organize recurring campaign aesthetics.
Midjourney produces visually polished results, but consistent identity across large image sets requires repeated reference use and manual selection. The service fits campaign ideation, editorial mockups, and social creative where art direction matters more than batch throughput. Teams needing private asset governance, programmatic generation, or direct integration with content systems face meaningful workflow limits.
- +Style References maintain a consistent visual language across portrait concepts
- +Moodboards collect reusable campaign aesthetics and reference images
- +Web editing supports inpainting, zooming, panning, and targeted variations
- +Strong lighting, composition, hair, and wardrobe control through prompts
- –No official public API supports automated production workflows
- –Consistent facial identity requires manual reference management
- –Discord access adds workflow complexity for teams using the web editor
- –Privacy controls require deliberate configuration for sensitive client work
Creative agency art directors
Build campaign portrait concepts
Faster concept approval
Editorial design teams
Prototype magazine cover imagery
Clearer visual direction
Show 1 more scenario
Brand social teams
Create culturally specific campaign assets
More creative variations
Teams generate varied portraits with controlled hair, wardrobe, complexion, and composition for social content testing.
Best for: Fits when creative teams need polished Black female portrait concepts with detailed art direction and manual review.
Stable Diffusion
API-firstOpen-source text-to-image diffusion model capable of generating diverse ethnicities via prompt and LoRA fine-tuning.
Downloadable model weights enable local portrait pipelines with custom checkpoints and application-specific preprocessing.
Stable Diffusion supports text-to-image generation, image-to-image editing, inpainting, and ControlNet conditioning through its surrounding ecosystem. LoRA fine-tuning can adapt recurring characters, facial traits, clothing, and lighting when teams provide suitable training images. Local deployment gives production teams direct control over model files, image storage, and inference settings.
The main tradeoff is operational complexity because local use requires GPU provisioning, model selection, interface configuration, and quality testing. A creative studio producing recurring campaign portraits can use custom weights and reference workflows to reduce variation across image sets. Results still differ substantially between model releases, checkpoints, and community extensions.
- +Downloadable weights support local generation and controlled data handling.
- +LoRA fine-tuning adapts recurring characters, lighting, and wardrobe.
- +Image-to-image and inpainting support targeted portrait revisions.
- +Stability AI API supports application-based generation without local GPU infrastructure.
- –Local deployment requires GPU provisioning, model selection, and interface configuration.
- –Facial identity can drift across prompts without reference-control workflows.
- –Output quality differs across checkpoints and community extensions.
- –Safety filtering and access controls vary between local interfaces and hosted endpoints.
Creative production studios
Recurring campaign portrait variations
Consistent campaign variants
Privacy-sensitive design teams
Local portrait generation
Internal image workflows
Show 2 more scenarios
Software product teams
API-based portrait generation
Embedded image generation
Stability AI API adds generated portraits to editors, mockup tools, and content pipelines.
Character artists
Reference-guided pose revisions
Faster directed revisions
ControlNet conditioning transfers pose or composition guidance while preserving prompt-defined appearance.
Best for: Fits when teams need controllable portraits of Black women across local workflows and API calls.
Leonardo.Ai
SMBGenerative AI platform offering fine-tuned models and prompt-based generation of diverse subjects.
Character Reference in Image Guidance preserves a subject’s facial identity across new scenes and portrait variations.
Leonardo.Ai earns its #4 position by combining multiple image models with reference-guided controls for portraits of Black women. Image Guidance supports character, style, content, and pose references, while Canvas adds inpainting and outpainting for targeted edits. Leonardo.Ai also provides an API for automated generation and model selection in production workflows.
- +Character Reference supports recurring facial identity across portrait variations.
- +Image Guidance accepts reference images for composition, style, and subject direction.
- +Canvas provides inpainting and outpainting for localized portrait corrections.
- +API access supports programmatic image generation in production workflows.
- –Exact hair texture and facial traits still require careful prompt iteration.
- –Character consistency can weaken across major pose, wardrobe, or lighting changes.
- –Advanced editing requires learning separate Canvas and guidance controls.
Best for: Fits when creators need reference-guided portraits of Black women with repeatable identity and an API-backed production workflow.
SeaArt AI
specialistWeb-based Stable Diffusion interface hosting community-trained models for diverse character generation.
SeaArt’s model-sharing ecosystem combines public model files, creator pages, and reusable generation settings in one workspace.
SeaArt AI generates Black female portraits from text prompts and reference images, with a community model-sharing ecosystem as its main distinction. Users can combine model selection, prompt weighting, negative prompts, image-to-image generation, inpainting, background removal, and face restoration.
Pose and composition tools support iterative edits without rebuilding every image from scratch. Output quality varies across community models, while visible API and governance controls remain limited for production automation.
- +Inpainting and background removal repair localized portrait defects without regenerating the entire composition.
- +Reference-image tools support pose, composition, and identity iteration.
- +Prompt weighting and negative prompts control clothing, lighting, and anatomy.
- +Face restoration can correct common portrait defects after initial generation.
- –Model quality varies sharply across community files, producing inconsistent facial features and skin rendering.
- –Public workflows can expose inconsistent settings, making repeatable output harder.
- –The consumer interface provides limited visible controls for API provisioning and audit logging.
- –Pose, face, and background corrections often require separate editing passes.
Best for: Fits when creators need many community models and manual controls for iterative Black female portrait generation.
Civitai
specialistRepository of community-trained Stable Diffusion checkpoints and LoRA models for specific demographics.
Model cards with trigger-word guidance and community-tested prompts for specific portrait outcomes.
Civitai is a model and LoRA sharing site used to generate images with melanin-consistent skin tone for ebony black skin female portraits. The core capability centers on browsing, downloading, and testing diffusion checkpoints and LoRAs, then pairing them with prompts in an external renderer like ComfyUI or Automatic1111.
Asset metadata such as model type, tested prompts, trigger words, and compatibility notes reduce the guesswork when building an inference workflow. Civitai also supports community-driven refinement through comments, tags, and variant uploads that affect how quickly a specific look reaches repeatable results.
- +Large library of diffusion checkpoints and LoRAs for consistent portrait looks
- +Model cards include trigger words and tested prompt notes for faster iteration
- +Community comments highlight failure modes and workable settings
- +Multiple download formats help route assets into common image toolchains
- –Quality varies per upload, so consistency needs manual vetting
- –No built-in inference engine for ControlNet-style conditioning workflows
- –Metadata often lacks skin-tone specific evaluation notes for edge cases
- –High dependency on external tooling for batching and workflow automation
Best for: Fits when building repeatable ebony black skin female portrait generations using community LoRAs and external inference tooling.
Fooocus
specialistOffline Gradio-based interface optimized for Stable Diffusion XL with simplified prompt inputs.
Preset-driven Gradio interface combines image prompts, inpainting, and outpainting without node graphs.
Fooocus differs from node-based image interfaces through a preset-driven Gradio workflow that keeps model settings behind a smaller control surface. Its SDXL pipeline supports text prompts, image prompts, style presets, inpainting, outpainting, and LoRA loading for local generation.
Automatic prompt expansion can add descriptive detail without requiring complex prompt engineering. Dark-skinned Black women are possible through explicit prompts, but complexion, facial features, and identity consistency vary between seeds.
- +Preset-driven controls reduce setup compared with node-based interfaces.
- +Image Prompt accepts reference images for composition and visual guidance.
- +Inpainting and outpainting are available inside the same generation workflow.
- +Local execution keeps source images and prompts on the configured machine.
- –No official REST API supports production automation or queue management.
- –Limited controls exist for pose conditioning and multi-image identity consistency.
- –Dark-skin complexion and facial features can drift between seeds.
- –Local generation requires compatible hardware and substantial model storage.
Best for: Fits when creators need repeatable portraits with image prompting and inpainting without an automation API.
Tensor.art
specialistOnline platform hosting Stable Diffusion models with online generation and model fine-tuning tools.
Reusable public workflows preserve complete generation setups, letting users replicate a community creator’s portrait configuration.
Tensor.art centers image generation around a community model library with reusable workflows, giving portrait creators more model choice than a single-model interface. Users can combine text prompts, image-to-image generation, inpainting, upscaling, LoRA assets, and ControlNet conditioning.
For ebony and dark-skin female portraits, the catalog supports targeted prompt testing across different rendering styles. Output consistency depends on selecting compatible models and maintaining stable generation settings.
- +Large community catalog provides many portrait-focused models and LoRA assets.
- +Reusable workflows preserve prompts, samplers, models, and generation settings.
- +Inpainting and image-to-image tools support targeted facial, hair, and wardrobe corrections.
- +Model pages provide examples that help compare rendering styles before generation.
- –Community uploads vary in quality, documentation, and prompt compatibility.
- –Consistent facial identity across separate generations requires manual workflow control.
- –Model and workflow selection can overwhelm users seeking a fast single-step generator.
- –Publicly shared settings do not guarantee identical results across model revisions.
Best for: Fits when creators need broad model choice and reusable community workflows for controlled portrait generation.
NovelAI
specialistAI storytelling and image generation platform with configurable character attributes.
Director's Tools provide inpainting, image-to-image editing, and regional revisions inside the same generation workspace.
NovelAI generates anime-oriented portraits of dark-skinned women through tag-based prompts, model selection, and configurable image settings. Its interface includes image-to-image generation, inpainting, regional editing, seeds, and an Undesired Content field for reducing unwanted details. The anime-focused training style can produce consistent illustration results, but photorealistic skin texture, natural lighting, and exact facial identity remain limited.
- +Tag-based prompts provide precise control over complexion, hairstyle, clothing, pose, and camera framing.
- +Inpainting and regional editing support targeted corrections without regenerating the entire image.
- +Seeds and image variations help reproduce similar character designs across multiple outputs.
- +Anime-focused models render stylized facial features and hair with consistent visual conventions.
- –Anime-focused output limits photorealistic portraits and natural skin reflectance.
- –Exact face consistency across separate generations requires repeated prompt and seed adjustments.
- –Complex prompts can produce uneven hands, jewelry, hair details, and background objects.
- –No documented public API supports automated batch generation or external workflow orchestration.
Best for: Fits when anime-style portraits need granular tag prompts, regional edits, and repeatable seeds.
Adobe Firefly
enterpriseGenerative AI image tool trained on licensed Adobe Stock content with explicit diversity and inclusion guidelines.
Adobe workflow integration combines Firefly generation with Generative Fill, Generative Expand, and reference-based editing.
Adobe Firefly suits teams that need branded portrait concepts inside Adobe workflows, but its demographic controls remain limited. Text-to-image generation supports Black female subjects, varied hairstyles, clothing, poses, and environments.
Generative Fill, Generative Expand, reference images, and style controls support further image refinement. Adobe Express, Photoshop, Illustrator, and Firefly Services APIs extend production use, although custom model training and detailed identity control are unavailable.
- +Generates Black female portraits with adjustable age, hairstyle, clothing, pose, lighting, and setting.
- +Generative Fill and Generative Expand repair framing problems without leaving the Adobe workflow.
- +Style and structure references provide more control than prompt-only image generation.
- +Content Credentials identify Firefly-generated images in supported Adobe workflows.
- –Facial identity can drift across variations and does not provide dependable character consistency.
- –No custom model training or local checkpoint export limits specialist control.
- –Prompt interpretation can flatten distinct African and diasporic features into generic representations.
- –Fine control over hands, jewelry, hair texture, and skin lighting remains inconsistent.
Best for: Fits when Adobe users need quick Black female portrait concepts for campaigns, mockups, and social content.
How to Choose the Right ai ebony black skin female generator
This guide ranks RAWSHOT AI, Midjourney, Stable Diffusion, Leonardo.Ai, SeaArt AI, Civitai, Fooocus, Tensor.art, NovelAI, and Adobe Firefly for ebony Black female image generation. RAWSHOT AI leads the ranking with seven-step visual configuration, editable selections, saved Stacks, and more than 1,800 licence-free synthetic models.
The comparison separates automated catalogue production from manual creative control. Midjourney uses Style References and Moodboards, Leonardo.Ai provides Character Reference, Stable Diffusion supports local model weights, and Adobe Firefly connects generation with Generative Fill and Generative Expand.
What Is an AI Ebony Black Skin Female Generator?
An ai ebony black skin female generator creates portraits or fashion images of Black women from text prompts, reference images, selectable settings, or reusable model workflows. Core differences include facial identity retention, hair and skin rendering, pose direction, editing controls, and repeatability across generations.
RAWSHOT AI uses separate controls for the model, garments, styling, background, lighting, and composition instead of a free-text prompt. Stable Diffusion supports downloadable weights, custom checkpoints, local generation, and LoRA fine-tuning for teams that need deeper model and workflow control.
Features that determine skin tone consistency, identity retention, and repeatability
Ebony Black skin female image generation succeeds when the workflow keeps lighting uniformity and preserves facial identity across repeated outputs. This guide prioritizes tools that manage identity and appearance with repeatable controls rather than free-form prompting.
Repeatable configuration blocks and saved output states
RAWSHOT AI replaces a free-text input with a seven-step visual configuration and saves selections as Stacks for repeatable catalogue treatments. Tensor.art preserves reusable community workflows so prompts, samplers, models, and generation settings can be replicated across runs.
Identity guidance across scene changes
Leonardo.Ai uses Character Reference in Image Guidance to preserve facial identity across portrait variations with reference images. Midjourney adds Style References and Moodboards to preserve campaign-specific visual direction across repeated Black female portrait generations.
Local control via downloadable model weights and fine-tuning
Stable Diffusion supports downloadable model weights so teams can run local portrait pipelines with custom checkpoints and preprocessing. It also supports LoRA fine-tuning for recurring characters, lighting, and wardrobe patterns.
Reference image support for identity and composition iteration
SeaArt AI includes reference-image tools that support pose, composition, and identity iteration while pairing with inpainting and background removal. Fooocus accepts image prompts and reference images inside a preset-driven Gradio interface to guide composition and visual direction.
Community model libraries with trigger-word prompt guidance
Civitai provides model cards with trigger-word guidance and community-tested prompt notes to speed up repeatable portrait outcomes. SeaArt AI adds a model-sharing ecosystem with public model files and reusable generation settings inside a single workspace.
Editing controls for localized corrections without full regeneration
NovelAI includes Director's Tools that support inpainting, image-to-image editing, and regional revisions inside one workspace. SeaArt AI provides inpainting and background removal that repair localized portrait defects without regenerating the entire composition.
How to choose an ai ebony black skin female generator with the right control model
The selection fork should start with workflow control depth. Some tools constrain inputs into editable blocks, while others rely on manual reference management or local checkpoint workflows.
Choose blocked configuration when catalogue consistency matters
RAWSHOT AI replaces the empty text box with seven visual configuration blocks for product, garments, styling, background, lighting, and composition, so each batch shares the same structure. Saved Stacks keep those selections visible and editable for repeat listings and large catalogue runs.
Choose reference-guided identity when the face must carry across scenes
Leonardo.Ai uses Character Reference in Image Guidance to carry facial identity into new scenes and portrait variations using reference images. Midjourney relies on Style References and Moodboards for campaign visual direction, so facial identity stability needs manual reference management.
Choose local weights when data handling and checkpoint control are required
Stable Diffusion supports downloadable model weights and local pipelines so teams can select checkpoints and run application-specific preprocessing. It also supports LoRA fine-tuning for recurring characters and wardrobe patterns, which reduces drift across repeated portrait batches.
Choose workflow reuse when repeatability must match a team’s existing pipeline
Tensor.art preserves complete generation setups through reusable public workflows, which can replicate prompts, samplers, models, and generation settings. That approach fits teams that need consistent outputs across separate operators without rewriting prompting logic.
Choose interactive editing when corrections must stay localized
SeaArt AI and NovelAI both support inpainting and targeted edits, but SeaArt focuses on repairing localized portrait defects and background issues without rerendering the full composition. NovelAI provides Director's Tools for regional revisions and image-to-image editing inside the same workspace.
Choose model libraries when speed matters more than strict identity controls
Civitai accelerates repeatability by packaging diffusion checkpoints and LoRAs with model cards that include trigger words and tested prompt notes. This approach still requires manual vetting because quality varies by upload, which can affect skin rendering consistency.
Who benefits from these ai ebony black skin female generator controls
Different teams experience failure modes in different places, like facial identity drift, lighting variance, or inconsistent skin rendering between batches. The tools below map to those failure points using their specific control mechanisms.
Indie labels and DTC apparel operators
RAWSHOT AI is built around configurable product, garments, styling, background, lighting, and composition blocks and then saves them as Stacks for repeatable catalogue treatments.
Creative teams producing campaign portrait concepts
Midjourney provides Style References and Moodboards to preserve campaign-specific visual direction across repeated Black female portrait generations, with manual review to maintain consistent faces.
Teams running local or controlled inference pipelines
Stable Diffusion supports downloadable model weights and local generation, and it adds LoRA fine-tuning to keep recurring characters consistent across wardrobe and lighting changes.
Creators iterating portraits with reference images and inpainting
SeaArt AI combines reference-image tools with inpainting and background removal to fix defects without regenerating the entire image composition.
Studios standardizing repeatable workflows across operators
Tensor.art stores reusable community workflows that preserve prompts, samplers, models, and generation settings so different operators can run the same generation setup.
Common pitfalls that break ebony black skin rendering and face consistency
Face drift and lighting variance usually come from switching control structures mid-run. Another common failure is assuming community model quality and prompt notes will behave uniformly across checkpoint versions.
Using free-text improvisation when a repeatable batch structure is required
RAWSHOT AI avoids free-text improvisation by replacing it with seven selectable blocks and saved Stacks, which keeps product, lighting, and composition aligned across runs.
Assuming facial identity is automatic when scenes or wardrobe change
Leonardo.Ai can preserve facial identity with Character Reference, but consistency can weaken across major pose, wardrobe, or lighting shifts if reference direction is not maintained.
Expecting a consistent identity pipeline without reference management
Midjourney has no official public API for automated production workflows, so consistent facial identity depends on manual reference management across generations.
Treating community-uploaded models as uniformly reliable
Civitai model quality varies per upload, so repeatable ebony black skin female portrait results still require manual vetting of each checkpoint and its trigger-word guidance.
Trying to do regional corrections without a tool that supports localized revisions
NovelAI’s Director's Tools support regional inpainting and revisions, and SeaArt AI supports inpainting and background removal that repair localized portrait defects without regenerating the entire composition.
How We Selected and Ranked These Tools
We evaluated RAWSHOT AI, Midjourney, Stable Diffusion, Leonardo.Ai, SeaArt AI, Civitai, Fooocus, Tensor.art, NovelAI, and Adobe Firefly using feature coverage for skin and identity control, then ease of producing consistent portraits, then value for repeatable production runs. Features account for 40% of the score, ease accounts for 30%, and value accounts for 30%.
RAWSHOT AI separated itself with a seven-step visual configuration that stays editable across batches and with Saved Stacks that preserve each selection for repeatable catalogue treatments. RAWSHOT AI also scored around the category lead with an overall 9.1 And features and ease near the top, with explicit commercial rights and more than 1,800 licence-free synthetic models.
Frequently Asked Questions About ai ebony black skin female generator
Which AI ebony black skin female generator is suited to apparel catalogue production?
How do teams connect an AI ebony black skin female generator to an existing production workflow?
Which tools provide the strongest control over facial identity and reference images?
When does local generation make more sense than a hosted AI portrait tool?
What breaks if a creator relies on community models for ebony black skin female portraits?
How should teams handle access control and image privacy for generated portraits?
Where does an AI ebony black skin female generator fall short for photorealism?
Which generator works best for anime-style ebony black skin female characters?
Conclusion
After evaluating 10 tools, RAWSHOT AI stands out as our overall top pick — it scored highest across our combined criteria of features, ease of use, and value, which is why it sits at #1 in the rankings above.
Use the comparison table and detailed reviews above to validate the fit against your own requirements before committing to a tool.
Tools reviewed
Primary sources checked during evaluation.
Referenced in the comparison table and product reviews above.
Keep exploring
Comparing two specific tools?
Software Alternatives
See head-to-head software comparisons with feature breakdowns, pricing, and our recommendation for each use case.
Explore software alternatives→Need a personal recommendation?
Software Advisory Service
Skip months of vendor evaluation. Our analysts recommend the right tool for your business in 2–4 weeks.
Talk to an analyst →