Articles on: Help

Glossary: generative AI and Seelab terms


Seelab and generative AI vocabulary, from A to Z: AI image and video generation, custom model training, editing, composition and production at scale. Button names are written exactly as they appear in the interface. Each definition links to the article that covers the feature in detail.


Organization and sharing


  • Collection: a selection of images and videos shared with project members. See Collections.
  • Credits: the platform's unit of consumption. The cost depends on the model, the resolution and the duration, and is shown before each run. See Understanding credits. See also Seelab pricing.
  • Feed: the private history of your generations, image and video. See The generation feed.
  • Managed services: support from the Seelab team, from modeling your products to producing visuals and training your teams. See Contact the Seelab team. See also onboarding and support, training and production.
  • Organization: your company's space, grouping projects, members and credits. See Organizations.
  • Owner, Admin, Member: the three member roles. Owner and Admin can invite. See Roles.
  • Project: a team's or a brand's shared space: models, collections, workflows. See Projects.


Image creation


  • Aspect ratio: the width-to-height ratio of the image, such as 1:1, 3:4, 9:16 or 16:9. Available formats depend on the model. See Settings and formats.
  • Multi-references: combining several reference images in a single generation, giving each one a role in the prompt ("the bottle from image 2 in the center"). See Bringing several products into one image.
  • Prompt: the text that describes the visual to generate, up to 3,000 characters. See Prompt deep dive.
  • Quality: on some models such as GPT Image, the rendering level (low, normal, high), which changes the detail and the cost.
  • Reference image: an image that guides the generation (a product, a setting, a mood, a style). The maximum number depends on the model. See Create a visual from a reference image.
  • Resolution: the size of the generated image, 1k, 2k or 4k depending on the model.
  • Resource Hub: the image picker panel: imports, your models' datasets, collections, generations. See The Resource Hub.
  • Reuse image: set an image from the feed as a reference to iterate.


Seelab models and training


  • Character model (Portraits & Avatars): a Seelab model trained on a face or a character, mascot, virtual model or brand hero, who stays recognizable from one image to the next.
  • Dataset: a Seelab model's training images. See How to photoshoot for custom training.
  • Key images: the 3 dataset images you mark with a 👍 because they best sum up the product, the character or the style.
  • Market model: a general-purpose generation model available in Seelab, such as Nano Banana, GPT Image, Flux or SeeDream. See Image models in detail.
  • Product model (Products & Objects): a Seelab model trained on a product, which respects its shape, logo, label and materials in any setting. See Create and train a model. See also product modeling.
  • Retrain: train a model again with new images. See Managing a model.
  • Seelab model: a model trained on your products, characters or styles from a few images. One training works in both image and video. See What is a Seelab model?.
  • Style model (Styles & Photography): a Seelab model that learns a way of seeing (light, palette, grain, framing) and applies it to any subject. See Style models: teach Seelab your way of seeing.
  • Training brief: the text or PDF that comes with the training: what the model is for, what it must respect, what it must avoid. See The training brief.


Image editing


  • Blend (Fusion): the integration of a layer into the scene by the Composer, with the setting's light and shadows.
  • Clean: erase an element from an image. See Clean an image.
  • Composer: the tool that assembles several visuals as layers in a single scene, then harmonizes light, reflections and shadows. See Composer. See also the Seelab Composer.
  • Magnify: Seelab's creative upscaler, which enlarges an image up to 4K while adding details. See Upscale (Magnify) an image. See also the AI editor.
  • PSD export: exporting a Composer composition as a Photoshop file, layer by layer. See Composer: layers, blending and PSD export.
  • Reframe: adapt an image to another format by extending the setting (outpainting). See Reframe (outpainting).
  • Remove background: cut out a layer in one click in the Composer.
  • Replace: replace an element in an image, or insert your product with your Seelab model (inpainting). See Replace and modify an image.
  • Replace background: change the setting behind a product or a person while keeping the subject intact. See Replace an image background.


Video


  • AI Engine: the video generation model, such as Kling, Gemini Omni Flash, SeeDance or Minimax. See Video models in detail.
  • Camera Presets: predefined camera moves and styles. See Prompting a video.
  • Lip-sync: the match between lip movement and voice, when the engine generates sound along with the image.
  • Multi Refs: the mode where several images, such as a character, a product and a logo, are combined in the video. See Multi-refs: character, product and logo.
  • Seelab video model: your Seelab model used in video, to animate your product without altering its logo or label. See Animate your product with your video model.
  • Start / End Frames: the mode where a start image, and optionally an end image, guides the video.
  • Video editing: the Edit tab of the Video space, which modifies an existing video (product, character, setting, style) while keeping its motion. See Edit a video.


Workflows



Generative AI terms


  • AI content factory: a content production setup where generative AI produces visuals at scale, with brand models and repeatable workflows. This is what Seelab lets you build. See also the Seelab AI content factory.
  • Brand consistency: the ability to produce visuals that always respect the same product, the same character and the same art direction. That is the role of Seelab models.
  • Diffusion model: the family of models that generates an image by starting from random noise and refining it step by step, guided by the prompt.
  • E-commerce visual: a product image for a product page, a marketplace or a digital campaign: packshot, lifestyle shot, on-model. See Building a furniture catalog with Seelab. See also AI product page visuals.
  • Fine-tuning: adjusting a generation model on your own images so it learns a product, a face or a style. On Seelab, this is custom model training.
  • Generative AI: artificial intelligence models that can create images, videos or text from an instruction.
  • Image-to-image: generating a new image from an existing image used as a reference.
  • Image-to-video: animating a still image into a video, for example an approved product visual. See How to generate video with Seelab. See also product videos.
  • Inpainting: regenerating a specific area of an image while keeping the rest. On Seelab: Clean and Replace.
  • LoRA: a lightweight training technique that adds knowledge of a product, a character or a style to a generation model, without fully retraining it.
  • Outpainting: extending an image beyond its edges. On Seelab: Reframe.
  • Packshot: a photo of a product on its own, usually on a neutral background, used in e-commerce. See Fashion · Maison Ferrande: from packshot to on-model.
  • Product fidelity: the exact match between the generated product and the real one: shape, proportions, logo, text, materials. See Using your model in image and video.
  • Text-to-image: generating an image from text alone. See Create an image with a text prompt.
  • Text-to-video: generating a video from text alone.
  • Upscaling: enlarging an image by increasing its resolution. On Seelab: Magnify, which adds details up to 4K.


Learn more


Updated on: 12/11/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!