Higgsfield Wiki
A plain-English reference for Higgsfield AI — its image and video generators, the face-swap tools, and the models it runs on.
Higgsfield AI is a creative platform for generating images and video from prompts and photos. This wiki documents what each part of it does, which underlying model powers it, and what to expect — written for people evaluating it or already using it.
Start here
How the platform is organised
Higgsfield groups its tools into a few families, and the distinction matters because they behave differently:
- Generators — start from a text prompt (and optionally a reference image) and produce something new. This covers the image generator and the video generator.
- Apps — single-purpose tools that transform something you upload: face swap, headshot generation, upscaling, background removal.
- Model pages — surfaces for specific third-party models the platform runs, such as Seedance 2.0, Sora 2, Veo 3 and Kling.
- Plugins — integrations into existing editing software, including DaVinci Resolve and Photoshop.
Most confusion about Higgsfield comes from mixing these up. A generator invents a new frame from scratch; an app modifies a frame you supplied. If identity preservation matters to you, that difference is the whole story — see face swap.
A note on models
Higgsfield does not train the frontier models it exposes. It provides an interface, presets, and workflow around models built elsewhere — Seedance from ByteDance, Sora from OpenAI, Veo from Google, Kling from Kuaishou, Nano Banana from Google. This is normal for the category and worth knowing when comparing tools: two platforms offering "Sora 2" are offering the same underlying generator, and competing on interface, price, and workflow rather than on output quality.
Common questions
Does Higgsfield build its own AI models?
Not the frontier generation models. It runs models from ByteDance, OpenAI, Google and Kuaishou, and builds the interface, presets and workflow around them.