CLIP ViT-B/32 model
Embeddings are compatible with CLIP text embeddings and with Video Embedding (CLIP) vectors for cross-modal search.
CatalogLatency Synchronous; typically <1 s Async false Rate Limit Per API key Max Input One image per request
cn8/studio-embed-image
A single synchronous endpoint powered by CLIP ViT-B/32 (via sentence-transformers).
Learn more about Image Embedding (CLIP) ImagesyncInstant previewimageanalysisembedding
Try it
Image to embed
Result
The result shows up here
Upload your image on the left and hit Run.
Overview
A single synchronous endpoint powered by CLIP ViT-B/32 (via sentence-transformers). studio-embed-image fetches an image URL, loads it as RGB, encodes it with CLIP, and returns a 512-dimensional unit-normalised float vector. Use the vector for semantic image search, nearest-neighbour retrieval, similarity scoring, and multimodal RAG. Embeddings share a semantic space with CLIP text embeddings and with the Video Embedding (studio-embed-video) endpoint, so you can search across modalities.
Key capabilities
Image embedding
A single image URL is fetched, loaded as RGB, encoded by CLIP, and a 512-dimensional float vector is returned.
Synchronous response
The embedding is returned directly in the JSON body — no polling.
When to use it
Semantic image search
Embed a library of images and query with a text or image vector to find visually similar results.
Image similarity
Compute cosine similarity between two embeddings to measure visual similarity.
Multimodal RAG
Store image embeddings in a vector database; retrieve relevant images for RAG pipelines.
Input & output
input
Image URL
JSON body
output
512-dimensional float vector
JSON
Guides & tips
CLIP embedding basics
- CLIP (ViT-B/32) produces 512-dimensional float vectors normalised to unit length.
- Vectors from the same model are directly comparable with cosine similarity.
- Text CLIP embeddings (from an external CLIP text encoder) are compatible: use them to query an image vector index.
Storing and querying embeddings
- Store embeddings in a vector database (e.g. Qdrant, Pinecone, Weaviate, pgvector).
- At query time, encode the query image or text with CLIP, then run nearest-neighbour search.
Specs
Schema
Request body
imageUrlrequiredstringPublic URL of the source image (jpg, png, webp, etc.)
Pricing
Billed per image encoded.
| Service | Unit | Price |
|---|---|---|
| Image Embedding (CLIP) | image | 0.5 credits/image |
Synchronous — billed once per successful encode.
FAQ
Related models
Video Embedding (CLIP)
Embed video frames into the same CLIP space for cross-modal image/video search.
Text Embeddings
Text embedding in the same semantic space to enable cross-modal search.
Image Captioning
Caption images for text-based search alongside embeddings.
Image Background Remove / Replace
Remove the background before embedding to focus CLIP on the subject.