Show it the screen, don't describe it
Bee Comb, Buzz, and Hive accept images and video natively over the OpenAI-compatible endpoint — screenshots, scans, diagrams, and recordings in the same conversation as text.
Native input, not an OCR sidecar
Vision and video input on Bee are served natively by the models on Comb, Buzz, and Hive — the model attends to the pixels and frames themselves, not to a lossy text transcription bolted on in front. That's the difference between reading a chart and reading a caption about a chart.
Requests use the standard multimodal Chat Completions format, so existing OpenAI-compatible clients send media without new code paths.
What teams point it at
- →Bug reports: a screenshot of the broken UI or a short screen recording of the repro, straight into the conversation.
- →Documents that only exist as images: scans, photographed whiteboards, stamped forms.
- →Diagrams and architecture drawings — ask what a component does or what's missing.
- →Recorded demos and walkthroughs: ask what happened, in what order, on the actual footage.
- →Dashboards and charts: compare two states and ask what changed.
Multimodal plus the rest of the platform
Because vision and video ride the same tiers as tool calling and JSON mode, multimodal workflows can be structured end to end: extract fields from a scanned form into schema-valid JSON, or feed a recording's analysis into a tool call. Text documents flow through retrieval alongside — one conversation, every modality.
Availability, stated exactly
Image and video input are served on Bee Comb, Buzz, and Hive. These are live-verified serving capabilities on the current deployment — Bee's policy is that a capability appears on this page only when the backend actually serves it.
Customer data is not used to train the Bee base model. Prompts, documents, and outputs stay yours; processing is limited to serving, securing, and operating the service.
Frequently asked questions
- Which tiers accept image and video input?
- Bee Comb, Buzz, and Hive serve native image and video input. The free Cell tier and Brood are text-focused surfaces on the current deployment.
- How do I send an image or video through the API?
- Use the standard OpenAI-compatible multimodal message format — media parts alongside text parts in the same messages array. Existing multimodal client code works unchanged.
- Can it combine a video with my documents?
- Yes — the same conversation can include video, images, and retrieval over your uploaded documents, on the tiers that serve each capability.
- Does Bee generate images or video?
- This page covers analysis — image and video input. Bee's workspace also serves image generation; video generation is not a claimed capability here.
- Are uploaded media files used for training?
- No — customer data, including media, is not used to train the Bee base model.
Related
Start on the free tier
Bee Cell is free — no card. Scale to paid tiers, the API, or sovereign deployment when you are ready.