Civitai Scraper

Search and extract Civitai AI model listings, full model/download details, image gallery posts, creators, and tags — no account or API key needed.

Why use this actor

• One actor covers Civitai's whole public surface: search models, pull full download-ready details for specific models, browse the image gallery, and list creators/tags
• Full model detail includes every version's file list — direct download URLs, file hashes (SHA256, BLAKE3, AutoV2...), virus/pickle-scan results, and base model — everything you need to build a model catalog or mirror
• No account, login, or API key required — this actor talks to Civitai's own official public REST API
• Automatic retries on temporary network hiccups
• Robust pagination — automatically follows Civitai's cursor/page links up to the limit you set, instead of returning just one page
• Honest about limits: modelId image filtering is fixed to work correctly even though Civitai's own API silently returns nothing for it (see Known limits) — and generation prompt metadata is never available publicly, which is stated plainly rather than faked

How it works

Pick a mode:

1. models — search/browse AI models (checkpoints, LoRAs, embeddings, etc.) with filters for query text, tag, username, type, sort order, and time window.
2. model-detail — fetch full detail (every version, every file, every download URL) for one or more specific model IDs.
3. images — browse the image gallery, optionally filtered by model, model version, username, or post.
4. creators — list or search usernames.
5. tags — list or search tags.

Input

Search models:

{
  "mode": "models",
  "query": "pony",
  "types": ["Checkpoint"],
  "sort": "Most Downloaded",
  "period": "AllTime",
  "maxItems": 50
}

Model detail (by ID):

{
  "mode": "model-detail",
  "modelIds": ["4201", "257749"]
}

Images:

{
  "mode": "images",
  "modelId": 4201,
  "sort": "Most Reactions",
  "maxItems": 50
}

Creators / Tags:

{ "mode": "creators", "query": "civ", "maxItems": 20 }

{ "mode": "tags", "query": "anime", "maxItems": 20 }

Field	Type	Description
mode	string	"models" (default), "model-detail", "images", "creators", or "tags".
query	string	Search term. models mode searches names, creators mode searches usernames, tags mode searches tag names.
tag	string	models mode — exact tag filter.
username	string	models/images mode — exact creator filter.
types	array	models mode — filter by model type(s) (Checkpoint, LORA, TextualInversion, etc.).
sort	string	Sort order — value set differs by mode (see input schema for the full list).
period	string	models/images mode — Day/Week/Month/Year/AllTime.
nsfw	string	images mode — content-rating filter.
modelId	integer	model-detail mode: the model to fetch. images mode: only images generated with this model (auto-resolved to its versions — see Known limits).
modelIds	array	model-detail mode — multiple model IDs in one run.
modelVersionId	integer	images mode — only images from this exact model version (faster/more direct than modelId).
postId	integer	images mode — only images from this post.
maxItems	integer	Max results to return (models/images/creators/tags modes). Pagination is followed automatically up to this cap. Default 100, max 10,000.
maxConcurrency	integer	model-detail mode — how many model IDs to fetch in parallel. Default 4.
proxyConfiguration	object	Apify Proxy settings. Residential is on by default (not required for this endpoint — it's Civitai's own official public API — kept for production robustness).

Output

Models / model-detail mode — one record per model, full upstream shape (truncated here for readability):

{
  "_input": "4201",
  "_source": "S1-api",
  "_scrapedAt": "2026-08-25T19:03:40Z",
  "id": 4201,
  "name": "Realistic Vision V6.0 B1",
  "type": "Checkpoint",
  "stats": { "downloadCount": 2304987, "thumbsUpCount": 52805 },
  "creator": { "username": "SG_161222", "image": "https://image.civitai.com/..." },
  "tags": ["photorealistic", "anatomical", "realistic"],
  "modelVersions": [
    {
      "id": 501240,
      "name": "V5.1 Hyper (VAE)",
      "baseModel": "SD 1.5 Hyper",
      "files": [
        {
          "name": "realisticVisionV60B1_v51HyperVAE.safetensors",
          "sizeKB": 2082642.47,
          "hashes": { "SHA256": "F47E942AD4C30D863AD7F53CB60145FFCD2118845DFA705CE8BD6B42E90C4A13" },
          "downloadUrl": "https://civitai.com/api/download/models/501240?fileId=418901",
          "virusScanResult": "Success"
        }
      ]
    }
  ]
}

Images mode — one record per image:

{
  "_input": "4201",
  "_source": "S1-api",
  "_scrapedAt": "2026-08-25T19:06:38Z",
  "id": 12221839,
  "url": "https://image.civitai.com/.../original=true/....jpeg",
  "width": 1024,
  "height": 1536,
  "nsfwLevel": "None",
  "stats": { "likeCount": 346, "heartCount": 113 },
  "meta": null,
  "username": "SG_161222",
  "baseModel": "SD 1.5 Hyper"
}

Creators / Tags mode:

{ "_input": "civ", "_source": "S1-api", "_scrapedAt": "...", "username": "civitdiff", "modelCount": 1, "link": "https://civitai.com/api/v1/models?username=civitdiff" }

{ "_input": "anime", "_source": "S1-api", "_scrapedAt": "...", "name": "anime", "link": "https://civitai.com/api/v1/models?tag=anime" }

A record with _error instead of the fields above means that item could not be processed (for example, "_error": "model_not_found" for a model ID that doesn't exist).

Known limits

• meta (prompt, sampler, seed, steps, CFG scale) is always null. This is Civitai's own privacy choice on the public API, not a bug — it holds true even for images whose own internal flags say a prompt exists. There is no public, unauthenticated way to retrieve it.
• images mode filtered by modelId is automatically fixed. Civitai's own API silently returns zero results for /images?modelId=X even on the most popular models on the platform — a genuine upstream defect (images are actually indexed by model version, not model). This actor detects a modelId input and transparently resolves it to that model's version IDs first, so you always get correct results. If you already know the exact model version you want, pass modelVersionId directly for a faster, single-request lookup.