tencent-ailab/IP-Adapter
Source: https://github.com/tencent-ailab/ip-adapter
Structured fields
{
"archived": false,
"checked_at": "2026-10-17T00:00:00.000Z",
"usefulness": "Lightweight (about 22M parameters) image-prompt adapter for SD 1.5 and SDXL: use a reference image as a prompt for style or subject, with plus, plus-face, full-face and (experimental) FaceID variants. A common building block for reference-based consistency and combining with ControlNet.",
"source_type": "repository",
"availability": "available",
"dependencies": [
"CLIP image encoders (OpenCLIP ViT-H/bigG)",
"diffusers (pinned 0.22.1 in README install)"
],
"source_owner": "Tencent AI Lab",
"next_check_at": "2026-11-16T00:00:00.000Z",
"known_limitations": [
"Mostly SD1.5/SDXL era; no FLUX/DiT port covered here.",
"FaceID variants marked experimental by authors.",
"Not archived per page but no recent updates visible (last log entry Jan 2024)."
],
"upstream_updated_at": "2024-06-28T02:47:03.000Z"
}Evidence claims
- [sourced_fact] README: IP-Adapter adds image prompting to pretrained diffusion models with about 22M parameters; variants include plus, plus-face, full-face, SDXL versions and experimental FaceID/FaceID-Plus/PlusV2/Portrait. source
- [sourced_fact] Limitation stated: CLIP center-crops input, so non-square references lose content outside the center; newest release-log entry is 2024-01-19 (FaceID-Portrait). source
- [sourced_fact] Hugging Face weights repo h94/IP-Adapter lists Apache-2.0 and provides ViT-H and bigG encoder variants and SD1.5/SDXL adapters. source
- [sourced_fact] Release log says IP-Adapter was integrated into Diffusers on 2023-11-22. source
- [upstream_claim] Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.
Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| code | Apache-2.0 | permitted | permitted | permitted | unknown | Repo page labels Apache-2.0. |
| model_weights | Apache-2.0 | permitted | permitted | permitted | unknown | h94/IP-Adapter card lists Apache-2.0 for the mirrored weights. FaceID variants depend on insightface face models, whose terms were not checked here (InstantID's README states insightface models are non-commercial research only). |