Video
Describe a scene, object, action, person, or place and open the matching video moment with its timestamp.
Describe what you remember.
Search, index, and connect your video library. Find exact timestamps through video, audio, images, or transcripts.
Talk to usSearch modes
Four ways to search the same video library.
Pureframe AI lets you search from the signal and noise, a scene description, spoken audio, a reference image or an exact transcript phrase.
Audio
Find where a speaker mentions a topic, question, customer, plan, or decision inside long recordings.
Image
Upload a still frame, screenshot, or product image and find visually similar clips in your collections.
Transcript
Look up phrases from transcripts and jump back to the source video moment where they were spoken.
What is Multimodal Engine
The One-Stop-Shop For Visual Intelligence
Multimodal engine transforms video into searchable intelligence by combining frame extraction, audio transcription, and visual embeddings. Process videos once and search across text, images, and audio, all powered by our unified multimodal processing pipeline.
Upload
Videos sent to secure storage
Processing Pipeline
Our engine transforms video into searchable content.
Extract Frame
Extracts frames, colors, text, and objects.
Transcribe
Converts speech into timestamped text.
Embed
Maps visual and spoken context for search.
Index
Structures and indexes everything for search.
Query
Your videos are ready to search.
API Gateway
App
SDK
MCP
Secure Storage
Videos and frames stored safely.
Multimodal Search
Search by text, image, audio, or transcript.
Developer First
APIs, SDKs, and webhooks built for builders.
Real-Time Ready
Fast indexing and instant results.
For Developers
One API call to search 10,000 hours.
REST API, typed SDKs, MCP and Webhooks. Results come back as ranked segments with thumbnail and clip URLs, ready to embed in your product or hand to an agent as a tool.
- Sub-second query latency at library scale
- Zero GPUs to provision, ever
- Drop-in tool spec for LLM agents
~zsh
curl https://api.pureframe.ai/v1/search \
-H "Authorization: Bearer $KEY" \
-d query="crowd starts cheering" \
-d modes="video,transcript"
{
"data": [{
"video_id": "vid_8x2k",
"segments": [{ "timestamp_start": 5021.0,
"score": 0.97,
"video_url": "PRESIGNED_URL"
}]
}]
}