$ man higgsfield

Higgsfield AI Glossary: 60 Key Models, Tools & Video Terms

60 Higgsfield AI terms in plain English: models, modes, camera moves, credits and plans. Type what you need and expand the definition. Full guide: Higgsfield AI - complete guide

This page contains affiliate links to Higgsfield. I may earn a commission if you buy through them.

60 / 60 terms

AI Influencer

AI Influencer is Higgsfield's character studio for building virtual personas from 19 menu settings without writing text prompts.

Users select a character type across three intensity levels - Average, Bold, or Extreme - and can optionally upload a photo to retain specific facial features. Each generation costs 1.12 credits, outputs both a close-up portrait and a full-body shot, and connects directly to Genjutsu in the Motion tab or to Cinema Studio.

Practical application: Build a recurring virtual host in the Influencer builder, save the persona as an element, and animate it with trending movements in the Motion tab.

Official documentation - Verified 2026-10-24

API

The Higgsfield API is a standalone, self-serve developer platform providing programmatic access to more than 50 video and image generation models, including Kling, Seedance, Wan, Soul 2, and DoP.

Unlike the website subscription, it uses pay-as-you-go billing in US dollars with a prepaid balance that stops automatically at zero and expires one year after top-up. Requests run asynchronously via REST or Python and TypeScript SDKs, and failed generations are refunded automatically.

Practical application: Integrate the Higgsfield API into your own software application when you need automated, per-request video or image generation billed in dollars rather than plan credits.

Official documentation - Verified 2026-10-24

Ads Studio

Ads Studio is a commercial video and image workspace on Higgsfield tailored for building promotional clips, social media hooks, and product-focused creatives.

It streamlines ad production by combining template-driven formats, avatar or product inputs, and rapid variation tools in one interface. Marketers use it to adapt campaign visuals across different aspect ratios and target audiences without filming new footage.

Practical application: Use Ads Studio when you need to turn a single product photo and a short marketing hook into multiple social video variations for testing.

Official documentation - Verified 2026-10-24

Arena Zero

Arena Zero is a flagship AI-generated Original Series featured within the Higgsfield ecosystem rather than a generation model or camera preset.

It showcases how episodic storytelling, action choreography, and recurring characters can be produced using the platform's filmmaking tools. Viewers can browse Higgsfield Originals through the Community filters to study high-end narrative production made on the platform.

Practical application: Watch Arena Zero in the Higgsfield Originals section to see how multi-scene action and consistent character design hold together across an episodic AI series.

Official documentation - Verified 2026-10-24

Aspect ratio

Aspect ratio describes the proportional relationship between the width and height of an image or video frame.

On Higgsfield, supported ratios depend on the chosen model and tool, commonly including 9:16 for vertical social feeds, 16:9 for standard widescreen, 1:1 for square posts, and 21:9 for ultrawide cinema in models like Kling 4.0. Selecting the appropriate frame shape before generating prevents awkward cropping later in post-production.

Practical application: Choose 9:16 when creating Reels or Shorts in Content Factory, and switch to 16:9 or 21:9 when directing widescreen narrative scenes in Kling 4.0.

Official documentation - Verified 2026-10-24

Assets

Assets is the private cloud library inside a Higgsfield account where all uploaded references and generated images, videos, and edits are stored by default.

Tools across the web interface and connected Claude MCP skills like Katana and Genjutsu read from and write back to this same central repository. Files in Assets remain private until a creator explicitly chooses to publish them to Higgsfield Community as a Shot or Open Project.

Practical application: Organize your best character portraits and generated clips into folders inside Assets so you can quickly pull them into Cinema Studio or Katana edits.

Official documentation - Verified 2026-10-24

Camera movement

Camera movement refers to how the virtual camera shifts position, angle, or focal framing during a shot to guide viewer attention and convey mood.

In Higgsfield, you can control motion either by selecting dedicated presets - such as the 30+ movement options in Cinema Studio 4.0 - or by describing the intended trajectory in your text prompt. Because not every model exposes a separate camera menu, clear prompt phrasing remains a reliable fallback.

Practical application: Select a built-in movement preset in Cinema Studio 4.0, or write a clear instruction like 'slow tracking shot following the runner from the side' in standard text-to-video prompts.

Official documentation - Verified 2026-10-24

Canvas

Canvas is Higgsfield's interactive image editing workspace designed for refining generated stills or uploaded photos without regenerating the entire picture from scratch.

It combines localized editing functions such as Inpaint brush masking, Relight adjustments, and camera angle modifications on a visual board. Note that unlimited plan benefits on higgsfield.ai do not extend to Canvas operations, which consume standard credits.

Practical application: Open a generated character portrait in Canvas to fix a small clothing detail or adjust directional lighting before using that image as a video first frame.

Official documentation - Verified 2026-10-24

Character consistency

Character consistency is the ability of a generative workflow to keep a subject's facial structure, skin tone, hair, and body proportions recognizable across multiple shots, poses, and lighting conditions.

Without dedicated anchoring, AI models tend to alter a person's jawline or eyes every time a new prompt is run. Higgsfield addresses this challenge through trained Soul ID profiles, Soul Cast actors, AI Influencer elements, and multi-image reference inputs.

Practical application: Combine a trained Soul ID for still keyframes with Kling 4.0 reference inputs or Cinema Studio @-mentions to keep your protagonist recognizable across a multi-shot sequence.

Official documentation - Verified 2026-10-24

Cinema Studio

Cinema Studio (currently at version 4.0) is Higgsfield's flagship AI filmmaking environment built for directing complex scenes up to 30 seconds long.

Cinema Studio 4.0 accepts up to 50 reference images and provides granular controls for genre, decade era, edit tempo, 30+ camera movements, 4 camera types (Modern, 35mm, 8mm, DV), 5 optically modeled lenses, aperture, 50+ color palettes, lighting, and an 8-type Emotion Wheel. It also supports Forward/Backward Extend to continue or lead into an uploaded video clip.

Practical application: Use Cinema Studio 4.0 when you want to direct a narrative scene using explicit menu controls for lens glass, aperture depth, era grain, and character emotion rather than relying on long text prompts alone.

Official documentation - Verified 2026-10-24

Commercial use

Commercial use refers to the pricing feature that permits creators to use generated videos and images in paid advertisements, client deliverables, monetized channels, and commercial products.

According to Higgsfield's official pricing table and API documentation, commercial use is included with paid subscription tiers (Starter, Plus, Ultra 3000) and pay-as-you-go API generations, but is not included on the Free plan. Regardless of your plan, you must always own or hold valid permissions for any input photos, videos, music, logos, or human likenesses you upload.

Practical application: Ensure you are on an active paid subscription plan or using the billed API before generating assets intended for client campaigns or monetized social accounts, and only upload source media you have the right to use.

Official documentation - Verified 2026-10-24

Community

Higgsfield Community is the public showcase section of higgsfield.

ai where creators share Open Projects, individual Shots, and blog posts on public profiles. Organized into Discovery, Projects, Shots, and Creators tabs, it tracks a public Community rating that awards 5,000, 75,000, and 200,000 generation credits upon reaching the 100, 500, and 1,000 point milestones. It also hosts Higgsfield Grants, where published Open Projects featuring a video of at least one minute can apply for up to 200,000 credits.

Practical application: Browse the Shots and Projects feeds in Community to inspect how other creators structure their prompts and models, or publish your own completed 1-minute film as an Open Project to apply for a Higgsfield Grant.

Official documentation - Verified 2026-10-24

Content Factory

Content Factory is an automated Higgsfield tool that transforms a single 'What if.

..' question into a finished faceless video running from 15 seconds to 5 minutes across 1 to 4 episodes. It writes the scene-by-scene script, selects characters, generates visuals and voiceover, and edits the final output using 50+ presets or a TikTok/Instagram reference link for pacing. Available on higgsfield.ai, via Claude MCP (/content-factory), and in Supercomputer, its credit cost ranges from 15.75 credits (one 15-second episode) to 1,260 credits (four 5-minute episodes).

Practical application: Enter a clear, escalating 'What if...' question and choose a tone preset in Content Factory to produce vertical 9:16 faceless storytelling videos for Reels, TikTok, or Shorts.

Official documentation - Verified 2026-10-24

Credit Mode

A mode in which a generation consumes a stated number of credits.

The cost depends on the model and settings, while feature access still depends on the plan.

Practical application: Check the task cost and remaining allowance before confirming.

Official documentation - Verified 2026-10-24

Credits

Credits are the internal unit of currency used on higgsfield.

ai and connected MCP tools to pay for image, video, audio, and editing generations. Current paid plans provide 270 credits on Starter ($19/mo), 1,200 credits on Plus ($59/mo or $47/mo billed annually), and 3,000 credits on Ultra 3000 ($129/mo or $99/mo billed annually), while creators can also earn credits through Community rating milestones and Grants. Subscription credits do not roll over from month to month, and annual plan credits are distributed periodically.

Practical application: Track your remaining monthly credits in your account header and test complex prompts at shorter durations and 720p resolution before spending more credits on 30-second 1080p renders.

Official documentation - Verified 2026-10-24

Depth of field

Depth of field is a filmmaking and photography term describing the distance range within a scene that appears sharp and in focus.

A shallow depth of field blurs the background to isolate a character's face, while deep focus keeps everything sharp from the foreground to the distant horizon. In Cinema Studio 4.0, you can set this directly via the Aperture control (f/1.4 Wide Open, f/4 Moderate, or f/11 Deep Focus), whereas in other models you can guide it through prompt descriptions.

Practical application: Select f/1.4 Wide Open in Cinema Studio 4.0 or add 'shallow depth of field with soft background bokeh' to your text prompt when generating close-up character portraits.

Official documentation - Verified 2026-10-24

Dolly in

Dolly in is a physical camera movement where the entire camera moves forward toward the subject through space, rather than merely magnifying the image with a lens zoom.

Because the camera actually travels closer, perspective and spatial depth between foreground and background elements shift naturally, heightening emotional focus or dramatic tension. In AI video generation, you can invoke this motion via camera presets where supported or by describing the forward physical camera travel in your prompt.

Practical application: Describe a 'slow dolly in toward the character's eyes as the room grows quiet' in your video prompt to build dramatic intensity during an emotional scene.

Official documentation - Verified 2026-10-24

Earn

Higgsfield Earn is a creator monetization platform where users link and verify an Instagram or YouTube account using a unique verification code placed in their profile bio or channel description.

Once verified, creators can browse active campaign briefs, publish qualifying videos or posts, and track performance-based payouts in real time on their dashboard. Withdrawals are processed through a connected Garna account; payouts depend on meeting campaign requirements and post performance, with no guaranteed income.

Practical application: Link your Instagram or YouTube profile in the Higgsfield Earn dashboard, verify ownership with the provided code, and review active campaign checklists before producing a promotional clip.

Official documentation - Verified 2026-10-24

End frame

A still image that guides the intended final frame of a clip.

It helps plan the ending, but the generated result must be checked against the reference.

Practical application: Use a product image as the target ending, then inspect the composition and details of the output.

Official documentation - Verified 2026-10-24

Face Swap

Face Swap is an editing technique that replaces the facial features of a subject in an image or video with the likeness from a target reference photo while preserving the original head movement, lighting, and expression.

Within Higgsfield workflows, facial identity replacement can be handled through dedicated swap tools, AI Influencer photo uploads, or character replacement passes. Users must ensure they only swap faces using their own likeness or photos of individuals who have given explicit permission.

Practical application: Apply a permitted reference photo to replace the actor's face in a generated test clip when you want to keep the existing lighting and body performance.

Official documentation - Verified 2026-10-24

First frame

A first frame is the initial still image used to anchor the opening composition, lighting, character appearance, and aspect ratio of an image-to-video generation.

Instead of letting a video model guess how a scene looks from text alone, supplying a polished first frame locks the visual design from second zero. On Higgsfield, creators frequently generate a high-detail still in Soul 2.0 or Nano Banana Pro and pass it into Kling 4.0 or Cinema Studio as the opening frame.

Practical application: Generate and refine your opening scene as a still image first, then upload it as the first frame in Create Video so your prompt only needs to direct the subsequent motion.

Official documentation - Verified 2026-10-24

Genjutsu

Higgsfield Genjutsu is a video transformation tool that rebuilds selected parts of an uploaded 4 to 30 second clip (real or AI-generated) using up to 30 reference images and 30+ presets, outputting up to 1080p.

It features two core modes: Motion Transfer, which extracts the movement from the input video and applies it to a new character or scene, and Object Swap, which replaces a specific element like an outfit or product while keeping the rest of the shot unchanged. Full text prompts are optional, and official credit pricing for a 15-second clip is 40 credits at 480p, 104 credits at 720p, and 144 credits at 1080p.

Practical application: Upload a 15-second smartphone product video to Genjutsu Object Swap along with reference photos of a new item to create localized ad variations without reshooting.

Official documentation - Verified 2026-10-24

Higgsfield DoP

Higgsfield DoP (Director of Photography) is Higgsfield's proprietary video generation model focused on cinematic framing, camera movement, and director-style visual execution.

Available on the higgsfield.ai platform and through the developer API (where it is listed at a starting rate of $0.125 per generation), it is engineered to interpret camera trajectory and lighting cues effectively. It serves as a core engine for creators who prioritize dynamic cinematography in short video clips.

Practical application: Select the DoP model when generating shots that rely on pronounced camera choreography and atmospheric studio lighting.

Official documentation - Verified 2026-10-24

Image-to-video

Image-to-video (I2V) is a generation workflow where a still image serves as the visual foundation and the AI model animates the subject, environment, and camera over time.

Because the appearance of the character, wardrobe, and set is already established in the input picture, I2V typically delivers much tighter visual predictability than generating video from text alone. Nearly all major video models on Higgsfield, including Kling 4.0, Seedance 2.5, and Veo 3.1, support image-to-video modes.

Practical application: Use an approved image from Soul 2.0 as your image-to-video input and write a concise prompt focusing strictly on how the subject and camera should move.

Official documentation - Verified 2026-10-24

Inpaint

Inpaint is a localized image editing technique where you brush or mask a specific area of a picture and instruct the AI to regenerate only that selected region.

The surrounding unmasked pixels remain untouched, allowing you to fix awkward hands, swap accessories, remove background distractions, or alter facial expressions cleanly. On Higgsfield, Inpaint is a core tool inside the Canvas workspace.

Practical application: Mask a character's jacket in Canvas Inpaint and describe a new leather texture to update their wardrobe without changing their face or background.

Official documentation - Verified 2026-10-24

Katana

Katana is a Higgsfield editing skill for Claude that runs through the Higgsfield MCP connector using your existing account subscription and credits.

You provide a reference video edit (either an uploaded file or a library preset) plus up to 5 photos of your characters, products, or clothing, and Katana rebuilds the video matching the reference's shot structure, cut timing, and pacing. Results are delivered as MP4 files directly inside the Claude chat widget and saved automatically to your Higgsfield Assets.

Practical application: Type /katana in Claude with the Higgsfield MCP connector active, pick a pacing preset, and upload 3 product photos to recreate a fast-cut commercial edit in chat.

Official documentation - Verified 2026-10-24

Keyframe

A keyframe is a still image placed at a specific point in a video timeline to define the exact visual state or pose the scene must hit at that moment.

While earlier video models only accepted a first and last frame, Kling 4.0 on Higgsfield supports up to 10 keyframes across a single generation of up to 30 seconds. The model interpolates the motion and camera transitions between each consecutive keyframe based on your text prompt.

Practical application: Place 4 to 6 sequential keyframes in Kling 4.0's Frames mode to map out a complex 20-second action scene beat by beat, using the prompt to describe the movement between frames.

Official documentation - Verified 2026-10-24

Kling 4.0

Kling 4.0 is a multimodal video generation and editing model available on Higgsfield that produces clips from 3 to 30 seconds at 720p or 1080p in Auto, 21:9, 16:9, 9:16, and 1:1 aspect ratios.

It supports prompts up to 8,000 tokens, up to 10 keyframes, up to 15 combined references (including up to 10 images and 5 videos totaling 30 seconds), two-channel stereo audio with multilingual voices, and on-screen text in nine languages. Official credit costs on Higgsfield are 25 credits (5 sec 720p), 37.5 credits (5 sec 1080p), 150 credits (30 sec 720p), and 225 credits (30 sec 1080p).

Practical application: Select Kling 4.0 in Create Video when you need extended 15 to 30 second scenes, multi-keyframe choreography, or stereo dialogue with stable fast-action physics.

Official documentation - Verified 2026-10-24

Lip-sync

Lip-sync is an audio-visual synchronization process that animates a character's mouth, jaw, and lower facial muscles to match spoken dialogue or a voiceover track.

In AI video workflows, lip-sync can be generated natively during video synthesis by audio-capable models or applied as a post-processing pass onto an existing talking-head clip. Accurate mouth alignment is critical for believable AI presenter videos, localized ads, and dramatic close-ups.

Practical application: Pair a clean, noise-free voiceover track with a front-facing character portrait or video clip so the model can align mouth shapes accurately to every syllable.

Official documentation - Verified 2026-10-24

MCP

MCP (Model Context Protocol) is the integration standard Higgsfield uses to connect conversational AI agents like Claude directly to your Higgsfield account via the server URL https://mcp.

higgsfield.ai/mcp. It requires an active Higgsfield subscription and account sign-in rather than a developer API key, and it exposes specialized chat skills such as Katana, Content Factory, and Genjutsu. All generations triggered through MCP spend your existing plan credits with no extra markup and save outputs to your Higgsfield Assets.

Practical application: Connect the Higgsfield MCP URL in Claude Settings under Connectors so you can generate Content Factory episodes or Katana video edits directly through chat prompts.

Official documentation - Verified 2026-10-24

Marketing Studio

Marketing Studio is a brand-oriented workspace on Higgsfield designed for generating e-commerce product photography, virtual lookbooks, and promotional campaign visuals.

It pairs with tools like Soul ID and product reference inputs so clothing labels and retailers can place items onto consistent models across studio or lifestyle settings. Its underlying image model, Marketing Studio Image, is also accessible via the Higgsfield API starting at $0.0059 per image.

Practical application: Use Marketing Studio together with a trained Soul ID to produce a cohesive seasonal fashion lookbook without booking a physical studio reshoot.

Official documentation - Verified 2026-10-24

Motion Control

Motion Control refers to directing the exact physical movement, gesture, or camera path in an AI video using a reference performance rather than text descriptions alone.

On Higgsfield, Kling Motion Control 3.0 lets you combine a clear character photo with a motion reference video to map body movements and facial expressions frame by frame at 720p or 1080p. Advanced settings also allow you to choose whether the background comes from the motion clip or the character image and customize the atmosphere via text prompt.

Practical application: Upload a clean dance or gesture clip alongside a full-body character image in Kling Motion Control 3.0 to transfer complex human choreography onto your custom avatar.

Official documentation - Verified 2026-10-24

Multi Reference

Multi Reference is a generation mode where you supply several distinct visual assets at once - such as a character face, a specific product, a clothing item, and a location photo - so the AI combines them into one coherent scene. In Higgsfield tools, reference capacity has expanded significantly: Kling 4.0 supports up to 15 references per generation, Genjutsu accepts up to 30 reference images, and Cinema Studio 4.0 accepts up to 50 reference images. Using multiple references reduces prompt guesswork and keeps recurring elements faithful to your source files.

Practical application: Upload separate reference photos for your actor, their vehicle, and the street environment so the video model assembles all three elements accurately in the same shot.

Official documentation - Verified 2026-10-24

Nano Banana Pro

Nano Banana Pro is a specialized image generation and editing model available within the Higgsfield suite, frequently used for fast concept art, graphic styling, and reference preparation.

Creators use it alongside Soul 2.0 and Seedream to produce sharp visual assets or modify existing compositions before moving into video animation. Its speed and prompt adherence make it a practical choice during early visual development.

Practical application: Generate stylized environment plates or prop concepts in Nano Banana Pro to use as reference inputs for your Cinema Studio or Kling video scenes.

Official documentation - Verified 2026-10-24

Negative prompt

A negative prompt is a supplementary text parameter that tells a generative model which visual elements, styles, or artifacts to exclude from the output.

Common exclusions include blurry textures, extra limbs, watermarks, distorted fingers, or unwanted color casts. While not every model exposes a separate negative prompt box in the user interface, specifying clean, affirmative visual descriptions in the main prompt helps avoid common artifacts across all models.

Practical application: When a model provides a negative prompt field, list specific unwanted traits like 'blurry, low contrast, oversaturated, extra fingers, text overlay' to keep the frame clean.

Official documentation - Verified 2026-10-24

Omni Reference

Omni Reference is a multimodal referencing system supported by models such as Kling 3.

0 and Kling 4.0 on Higgsfield that allows users to anchor characters, objects, scenes, and styles across generations. In Kling 4.0, Omni Reference works alongside expanded multimodal video editing, accepting up to 15 combined references (up to 10 images and 5 videos) to maintain continuity or rework existing clips. It enables targeted changes to a character, background, or visual style while preserving the underlying scene structure.

Practical application: Use Omni Reference in Kling 4.0 to lock your hero character and product bottle while generating multiple angles of the same commercial sequence.

Official documentation - Verified 2026-10-24

Open Project

An Open Project is a comprehensive portfolio entry published to Higgsfield Community that bundles a finished video or image series together with its project brief, asset folders, generation count, and models used.

During the four-step publishing flow (Details, Assets, Cover, Access), creators choose either 'With prompts' to let viewers inspect every prompt or 'Prompts hidden' to keep prompt text private while sharing the folders and result. An Open Project with a video, at least 30 likes, and 30 generations is automatically sent for verification review (+30 rating points), and projects with a 1-minute+ video can apply for Higgsfield Grants up to 200,000 credits.

Practical application: Publish your completed short film as an Open Project with organized asset folders to build your Community rating and qualify for Higgsfield Grants.

Official documentation - Verified 2026-10-24

Orbit

Orbit (also called an arc shot) is a camera movement where the camera travels in a circular or semi-circular path around a central subject while keeping that subject framed in the center.

As the camera circles, the background sweeps rapidly across the frame, creating a strong sense of three-dimensional space and heroic emphasis. In AI video workflows, you can trigger an orbit via camera movement presets in workspaces like Cinema Studio 4.0 or by describing the circular path in your prompt.

Practical application: Choose an orbit preset or prompt a 'smooth 180-degree orbit camera movement around the subject standing in the rain' to create a dramatic hero shot.

Official documentation - Verified 2026-10-24

Pan

A pan is a horizontal camera rotation where the camera turns left or right from a fixed position, similar to turning your head from side to side without walking.

Unlike a lateral tracking or truck shot where the camera body moves sideways through space, a pan pivots on a stationary axis to reveal a wider landscape or follow a moving subject. Cinema Studio 4.0 includes explicit presets such as Pan Left, and any text-to-video model can be guided by specifying 'pan left' or 'pan right' in the prompt.

Practical application: Use a slow pan right in your prompt or camera menu to reveal a sprawling city skyline starting from a character standing on a balcony.

Official documentation - Verified 2026-10-24

Popcorn

Popcorn is Higgsfield's storyboard and multi-panel visual planning tool designed to turn a script or scene concept into a connected sequence of still frames before committing to full video generation.

By mapping out shot framing, character blocking, and narrative flow in still panels first, creators can spot continuity issues early at a fraction of the credit cost of rendering multiple video takes. Approved Popcorn panels can then be exported or used as first frames and keyframes in video models.

Practical application: Draft your 6-shot commercial concept in Popcorn first to verify the visual story arc, then feed the best panels into Kling 4.0 or Cinema Studio as keyframes.

Official documentation - Verified 2026-10-24

Preset

A preset on Higgsfield is a pre-configured combination of visual style, lighting, motion, or editing parameters that applies a specific look or format in one click without complex manual prompting.

Presets appear across the entire suite: Soul 2.0 offers 20+ photography presets (like Flash Editorial or Y2K Studio), Cinema Studio 4.0 provides 30+ camera movement and 50+ color palette presets, Content Factory includes 50+ story format presets, and Katana and Genjutsu host libraries of motion and edit presets. Using a tuned preset often produces more predictable stylistic results than freehand text descriptions.

Practical application: Select a ready-made preset in Soul 2.0 or Genjutsu to establish the aesthetic baseline automatically, adding only a brief prompt for subject-specific details.

Official documentation - Verified 2026-10-24

Product Placement

Product Placement in generative AI is the workflow of inserting a real commercial item - such as a beverage can, cosmetic bottle, gadget, or garment - into a generated or existing video scene with natural lighting and perspective. On Higgsfield, creators achieve this through Genjutsu Object Swap, Katana reference edits, Marketing Studio, or multi-reference video modes in Kling 4.0 and Cinema Studio 4.0. Providing clear, well-lit reference photos of the product from multiple angles helps preserve label geometry and brand colors.

Practical application: Upload clean front and side photos of your product into Genjutsu Object Swap or Katana to place the item naturally into an existing commercial video structure.

Official documentation - Verified 2026-10-24

Prompt

A prompt is the written instruction given to an AI model to describe the subject, action, environment, lighting, camera movement, and audio of a desired generation.

On Higgsfield, prompt requirements vary widely by tool: Kling 4.0 accepts detailed multi-shot descriptions up to 8,000 tokens, whereas AI Influencer Studio requires zero prompt writing and Genjutsu can run entirely on presets with optional short text notes. When combining a prompt with keyframes or reference images, the text works best when it focuses on action and transitions rather than re-describing static visual traits.

Practical application: Structure your video prompt in clear order - subject action first, environment and lighting second, and camera movement last - while letting reference images handle facial details.

Official documentation - Verified 2026-10-24

Reference image

A reference image is an uploaded or previously generated still picture passed into an AI tool to guide the identity of a person, the shape of a product, the outfit, or the overall artistic style.

Unlike a first frame, which forces the video to begin with that exact composition, a reference image lets the model place that subject or style into brand-new angles and poses. Higgsfield tools support extensive reference inputs, including up to 5 photos in Katana, up to 10 images (plus 5 videos) in Kling 4.0, up to 30 in Genjutsu, and up to 50 in Cinema Studio 4.0.

Practical application: Supply 3 to 5 clear, unobstructed reference images of your character and prop when using Kling 4.0 References mode so the model maintains visual continuity across new camera angles.

Official documentation - Verified 2026-10-24

Resolution

Resolution measures the pixel dimensions and sharpness of a generated image or video, commonly expressed on Higgsfield as 480p, 720p, or 1080p for video, and up to 4K for certain still image workflows.

Higher resolutions deliver crisper detail for commercial and widescreen viewing but consume substantially more credits per second of output. For example, a 5-second Kling 4.0 clip costs 25 credits at 720p compared to 37.5 credits at 1080p, and a 15-second Genjutsu clip scales from 40 credits at 480p to 104 at 720p and 144 at 1080p.

Practical application: Draft and test your motion concepts at 480p or 720p to conserve credits, and switch to 1080p only when rendering the final approved take.

Official documentation - Verified 2026-10-24

Seedance 2.5

Seedance 2.5 is a fast, cost-efficient video generation model from the Seedance family available on the Higgsfield platform and through the Higgsfield developer API.

On the API price list, Seedance 2.5 starts at $0.0738 per second of output, positioning it as a much lighter-weight option than the heavier Seedance 2.0 tier ($0.9332 per second). Creators use it for fluid character motion, dance sequences, and everyday image-to-video tasks where speed and budget matter.

Practical application: Choose Seedance 2.5 when you want to iterate rapidly on expressive character movement or social clips at a moderate generation cost.

Official documentation - Verified 2026-10-24

Seedream

Seedream is a high-resolution text-to-image and image-editing model family integrated into the Higgsfield creative suite alongside Soul 2.

0 and Nano Banana Pro. It is recognized for strong prompt comprehension, artistic versatility, and clean rendering of complex lighting and textures across both photorealistic and stylized compositions. Images created with Seedream can be saved directly to Assets and passed into video models as first frames or references.

Practical application: Use Seedream to generate richly textured concept art or stylized key visuals when your project calls for a less strictly photographic look than Soul 2.0.

Official documentation - Verified 2026-10-24

Shot

In general filmmaking, a shot is a single continuous take recorded without an edit cut; in Higgsfield Community, a Shot is also a specific public publishing format for a standalone image or video.

When a creator publishes a Shot from their Assets library to the Community Shots feed, it carries the media file along with its exact prompt, model name, generation settings, and a one-click recreate option. This allows other users to study the configuration and test the setup with their own variations.

Practical application: Publish a standout single video or portrait from your Assets as a Community Shot to showcase your prompt craft on your public profile.

Official documentation - Verified 2026-10-24

Soul

Soul (currently Soul 2.0, along with Soul Cinema) is Higgsfield's proprietary photorealistic AI image model engineered for natural skin textures, authentic lighting, and editorial photography aesthetics.

It ships with more than 20 built-in style presets - such as Warm Ambient, Flash Editorial, Street Photography, and Y2K Studio - and is also accessible on the developer API at $0.0032 per image. Unlike Soul ID (an identity layer) or Soul Cast (an actor builder in Cinema Studio), Soul 2.0 is the core image generation model itself.

Practical application: Select Soul 2.0 with an editorial preset to generate lifelike portraits or fashion stills that avoid the waxy, over-smoothed look common in generic image models.

Official documentation - Verified 2026-10-24

Soul ID

Soul ID is Higgsfield's trainable character identity layer that locks a real person's facial structure and proportions for reuse across Soul 2.

0 image generations. To train a Soul ID, you upload at least 20 well-lit, varied-angle photos of the same person (including one full-height shot and no sunglasses), and training completes in about 3 to 5 minutes. Unlike Soul Cast, which builds fictional video actors from parameter menus inside Cinema Studio, Soul ID learns a specific real likeness from uploaded photos and applies it across Soul 2.0 presets without re-uploading references.

Practical application: Train a Soul ID once with 20+ recent photos of yourself or a consented model so you can generate dozens of consistent personal branding or lookbook images in Soul 2.0.

Official documentation - Verified 2026-10-24

Start & End Frame

Start & End Frame is a dual-anchoring video generation control where the user supplies both the opening still image (first frame) and the closing still image (end frame) of a clip.

The video model generates the intermediate motion to bridge the visual gap between the two images over the selected duration. This technique is useful for controlled transformations, morphs, camera push-ins between two known angles, or seamless loop creation.

Practical application: Provide a wide establishing shot as the start frame and a medium close-up of the same character as the end frame to guide a controlled camera transition.

Official documentation - Verified 2026-10-24

Storyboard

A storyboard is a visual sequence of ordered still panels that maps out the framing, action, and continuity of a video project before motion rendering begins.

Within the Higgsfield workflow, storyboarding can be handled automatically by tools like Popcorn or assembled manually by generating consistent stills in Soul 2.0 and arranging them as keyframes. Planning via a storyboard prevents costly trial-and-error in long video generations.

Practical application: Build a 4 to 8 panel storyboard before starting a 30-second Cinema Studio or Kling 4.0 project so you can verify character outfits and camera angles at low image credit cost.

Official documentation - Verified 2026-10-24

Supercomputer

Higgsfield Supercomputer is an agentic, chat-driven workspace on the platform that orchestrates multi-step creative tasks using modular skills such as Content Factory.

In Supercomputer workflows, complex jobs like scripting scenes, generating separate per-scene voiceover lines, creating visuals, and assembling a final edit can be managed conversationally with individual scene regeneration. Note that unlimited web generation perks do not apply to Supercomputer, which deducts standard plan credits.

Practical application: Run Content Factory as a skill inside Supercomputer when you want to review a multi-scene 'What if' script and regenerate a single voiceover line without redoing the entire video.

Official documentation - Verified 2026-10-24

Text-to-image

Text-to-image (T2I) is the process of generating a brand-new still picture from a written text prompt and optional style presets.

On Higgsfield, creators use text-to-image models like Soul 2.0, Seedream, Nano Banana Pro, and Marketing Studio Image to design characters, environments, and product shots from scratch. Because still images cost only a fraction of a video clip, T2I serves as the primary ideation and keyframe-building stage for most video pipelines.

Practical application: Start a new project in text-to-image mode to explore lighting and wardrobe ideas cheaply before animating your favorite still in a video model.

Official documentation - Verified 2026-10-24

Text-to-video

Text-to-video (T2V) is a generation mode where a video clip is synthesized directly from a written prompt without requiring an initial input image.

The model invents the visual composition, characters, lighting, motion, and audio simultaneously based on the text instructions and any selected menu settings. While T2V is fast for brainstorming standalone scenes or b-roll in models like Kling 4.0 or Veo 3.1, pairing it with reference images or keyframes is recommended when strict character continuity is required.

Practical application: Use text-to-video to generate atmospheric b-roll, establishing shots, or one-off action ideas when you do not yet have a reference image.

Official documentation - Verified 2026-10-24

Tilt

A tilt is a vertical camera rotation where a stationary camera angles upward (tilt up) or downward (tilt down) on a fixed pivot, much like nodding your head up or down.

It differs from a pedestal or crane movement, where the entire camera body rises or lowers vertically through space. You can direct a tilt in Higgsfield either through camera movement presets in Cinema Studio 4.0 or by writing 'camera tilts up' or 'camera tilts down' in your video prompt.

Practical application: Prompt a 'slow camera tilt up from the character's muddy boots to their determined expression' to introduce a hero or reveal a towering building.

Official documentation - Verified 2026-10-24

Unlimited

Unlimited refers to a subscription plan feature on higgsfield.

ai that allows paid users to generate outputs on eligible models without deducting from their monthly fast-credit pool. According to official platform restrictions, Unlimited/free generations are strictly limited to the higgsfield.ai web interface and never apply to MCP, CLI, Canvas, or Supercomputer workflows. Furthermore, newly launched models roll out gradually and are not automatically included in Unlimited mode.

Practical application: Use eligible Unlimited web modes for high-volume brainstorming on supported models, while reserving your monthly credit pool for Kling 4.0, Canvas edits, and MCP tools.

Official documentation - Verified 2026-10-24

Upscale

Upscale is a post-processing operation that increases the pixel resolution and sharpens fine details of an existing image or video clip.

Instead of paying peak credit rates to render every experimental video take at maximum resolution, creators can generate drafts at 480p or 720p and upscale only the final selected clip. Upscaling helps prepare AI-generated visuals for high-definition displays and editing timelines.

Practical application: Apply Upscale to your final approved video cut or hero image so it blends cleanly into a 1080p or 4K post-production timeline.

Official documentation - Verified 2026-10-24

Veo 3.1

Veo 3.1 is a high-definition third-party video generation model integrated into the Higgsfield multi-model suite alongside Kling and Seedance.

Known for strong prompt adherence, natural environmental physics, and cinematic lighting, it is used by creators for polished b-roll, commercial shots, and narrative sequences. Because premium video models consume credits at higher rates per second, comparing Veo 3.1 costs against Kling 4.0 or Seedance 2.5 before generating is recommended.

Practical application: Select Veo 3.1 in the video model picker when you need realistic environmental physics and cinematic lighting for a hero commercial shot.

Official documentation - Verified 2026-10-24

Voiceover

A voiceover is a spoken narration or dialogue audio track generated to accompany visuals in an AI video.

On Higgsfield, automated voiceover generation is built directly into tools like Content Factory (where each scene's line can be regenerated individually in Claude and Supercomputer) and supported natively by multimodal models like Kling 4.0 across multiple languages and accents. In Content Factory via Claude, the narration uses the language of your prompt, though languages other than English and Russian may carry an accent.

Practical application: Use Content Factory in Claude or Supercomputer to generate scene-by-scene narration automatically, regenerating individual voiceover lines if you tweak the script.

Official documentation - Verified 2026-10-24

Higgsfield AI Glossary: 60 Key Models, Tools & Video Terms

Content

Models

Prices and data

Trackers and law

Learn

Tools and more