Use Cases

Internal footage retrieval

Problem: Teams sit on hours of raw recordings — sales calls, user interviews, training sessions — and finding a specific moment means scrubbing through video manually.

How Pureframe helps: Upload recordings to a collection and search with natural language. "customer mentions pricing" returns the exact timestamp across hundreds of hours of footage in under a second.

$
curl -X POST https://api.pureframe.ai/v1/search \
>
  -H "Authorization: Bearer pf_key_..." \
>
  -F "query=customer mentions pricing" \
>
  -F "collection_id=col_sales_calls"

AI agent with video context

Problem: AI assistants can’t see video. When users ask questions about recorded content, the agent has no way to look it up.

How Pureframe helps: Connect the Pureframe MCP server to Claude, GPT-4o, or any LLM with function calling. The agent can search footage, retrieve frame images, and reason about visual content in a single conversation turn.

Example agent prompt: "Find the part of last week's demo where we showed the pricing page, and tell me what the presenter said about the annual plan."

The agent searches, retrieves the frame, reads the transcript, and answers — without leaving the chat interface.

Video search for end users

Problem: You’re building a product that includes video content and users need to find specific moments without scrubbing.

How Pureframe helps: Use the search API to power a search bar in your app. Users type what they’re looking for, and your UI displays the matched clips with thumbnails and timestamps. Results include presigned playback URLs ready for your video player.

Content moderation at scale

Problem: You have a library of user-generated video and need to identify specific types of content (e.g., sensitive scenes, brand mentions, competitor references).

How Pureframe helps: Run recurring searches across your library using the modes=["transcript"] and modes=["video"] parameters to surface content that matches your criteria. Automate flagging and review workflows using the API.

Training data curation

Problem: You need labeled video clips for a fine-tuning or evaluation dataset, but manually reviewing hours of footage is not practical.

How Pureframe helps: Describe the moments you need — "person looking directly at the camera", "close-up of hands on keyboard", "outdoor scene with natural light" — and extract matched clips with timestamps. Use thumbnail_base64 to pass frames directly to a vision model for automated labeling.

Meeting and lecture search

Problem: You record meetings or lectures and want attendees to find specific topics discussed without watching the whole recording.

How Pureframe helps: Upload recordings to a per-meeting or per-course collection. Expose a search box powered by POST /v1/search with modes=["transcript"] — users search what was said and jump directly to the moment in the recording.