Understand images and audio
Route incoming attachments, use Message ID for analysis or transcription, and handle processing outcomes explicitly.
- TRIGGERReceive Message
- STEP 01Analyze Image or Transcribe Audio
- STEP 02Use text or handle the fallback
Receive the right message type #
Add Receive Message to a conversation workflow. Enable the types you expect: text, audio, image, video, or file. The node’s branch options follow those enabled types.
Use the Image branch for image analysis and the Audio branch for transcription. A generic file is not automatically an image or audio input.
Analyze an image #
Add Analyze Image, select an available native OpenAI connection, and insert the received Message ID using the variable picker. Choose an appropriate model in the inspector and provide concise instructions about what to look for.
For example, ask for the product name visible on a package rather than a broad description if that is all your workflow needs.
Transcribe audio #
Add Transcribe Audio to the Audio branch, select a compatible connection and model, and pass the Message ID. Optional transcription context can help with domain-specific terms.
The incoming-audio processing path includes Telegram voice-note formats such as OGG/Opus. Check the processing result rather than assuming every provider or file is accepted.
Handle success and pending/error outcomes #
These nodes expose Success and Error / pending paths. Check text, transcript, response, error, and status outputs as appropriate for the action. A pending result is not completed text ready to send to a customer.
Give a clear fallback, such as asking the customer to type the information or letting the team follow up. Avoid repeatedly submitting the same attachment while waiting.
Understand account use #
Processing sends attachment content to the selected provider and can incur charges. Use only attachments and connected accounts appropriate for the task. Repeating identical processing settings can reuse the saved result; a different model or instruction can create different processing work.
For processing an external image reference rather than an incoming conversation attachment, see the relevant OpenAI node. Select the connection compatible with that specific action.