Media Understanding
Anyy can use attached media as conversation input when a compatible understanding route is configured. Image, video, and audio routes are independent; configuring one does not automatically enable the others.
Attach Media in Dashboard
Open Dashboard → Chat, choose or create a session, and attach files before sending the message. A single upload request accepts up to five files, each up to 50 MiB. Pending attachments expire after 24 hours if they are not claimed by a message.
Give an explicit instruction with the attachment:
Read this screenshot and list the visible error messages in order.
Summarize the main transitions in this video and note the timestamps that need review.
Supported Understanding Inputs
| Medium | Runtime input |
|---|---|
| Image | JPEG or PNG attachment for image understanding |
| Video | MP4, WebM, or QuickTime attachment for video understanding |
| Audio | Transcript-first processing through the configured audio/STT route |
The Dashboard can store other files as attachments, but model understanding depends on a supported media type and an enabled route.
Configure the Routes
Use Dashboard → Model to configure the main model and auxiliary image, video, or speech routes. Anyy validates that a selected provider/model declares the capability required by its slot.
See Models & Providers for provider authentication, fallback order, and auxiliary model setup.
Processing Modes
Image and video understanding can be configured to choose a native-capable route first or produce a bounded summary first. Audio understanding uses transcript-first processing. Failure behavior is configurable; it can stop the request or annotate/ignore a failed audio transcription according to the selected policy.
You normally do not need to choose a mode per message—the Profile configuration defines the route.
For images, ask for visible text, layout, or a comparison. For video, request timestamps or a sequence. For audio, specify whether you want a verbatim transcript, summary, speakers, or action items.
Privacy and Cost
Media bytes may be sent to the configured provider. Review that provider's retention and regional terms before uploading sensitive content. Large images, long videos, and audio transcription can increase latency, token use, and provider cost.
Troubleshooting
| Problem | Check |
|---|---|
| Attachment uploads but is not understood | Matching media route is enabled and model capability is declared |
| Image is rejected | Use JPEG or PNG and stay under the upload limit |
| Video is rejected | Use MP4, WebM, or QuickTime with a compatible video model |
| Audio has no transcript | Configure and test the audio/STT route |
| Provider returns a size/context error | Compress or trim the media, or choose a compatible provider |