Files Audio

Dify — Files & Audio

Uploading files for multimodal messages, and the speech-to-text / text-to-speech endpoints. For auth/base URL/user see the setup skill.

POST /files/upload — upload a file

Upload once, then reference the returned id as upload_file_id in a files array on /chat-messages, /completion-messages, or /workflows/run (with transfer_method: "local_file"). Multipart form; the file is scoped to user.

curl -X POST 'https://api.dify.ai/v1/files/upload' \
  --header 'Authorization: Bearer app-XXXX' \
  --form 'file=@/path/to/image.png' \
  --form 'user=user-123'

{
  "id": "...",
  "name": "image.png",
  "size": 204800,
  "extension": "png",
  "mime_type": "image/png",
  "created_by": "...",
  "created_at": 1700000000
}

Then use it:

curl -X POST 'https://api.dify.ai/v1/chat-messages' \
  --header 'Authorization: Bearer app-XXXX' \
  --header 'Content-Type: application/json' \
  --data '{
    "query": "Describe this image",
    "inputs": {},
    "response_mode": "blocking",
    "user": "user-123",
    "files": [
      { "type": "image", "transfer_method": "local_file", "upload_file_id": "<id-from-upload>" }
    ]
  }'

Allowed extensions/sizes depend on the app's file_upload config — check GET /parameters (see setup skill). type is image / document / audio / video.

GET /files/{file_id}/preview — preview / download an uploaded file

curl 'https://api.dify.ai/v1/files/{file_id}/preview' \
  --header 'Authorization: Bearer app-XXXX' \
  --output downloaded-file

Returns the binary content. Availability depends on Dify version/storage configuration.

POST /audio-to-text — speech to text

Transcribe an audio file. Multipart form.

curl -X POST 'https://api.dify.ai/v1/audio-to-text' \
  --header 'Authorization: Bearer app-XXXX' \
  --form 'file=@/path/to/audio.mp3' \
  --form 'user=user-123'

{ "text": "transcribed speech here" }

Supported formats: mp3, mp4, mpeg, mpga, m4a, wav, webm (≤ ~25 MB). Requires a speech-to-text model configured on the app.

POST /text-to-audio — text to speech

Generate audio from text (or from an existing message_id).

curl -X POST 'https://api.dify.ai/v1/text-to-audio' \
  --header 'Authorization: Bearer app-XXXX' \
  --header 'Content-Type: application/json' \
  --data '{ "text": "Hello from Dify", "user": "user-123" }' \
  --output speech.mp3

You may pass message_id instead of text to synthesize an existing answer. Requires a text-to-speech model configured on the app. Returns binary audio (audio/wav or audio/mp3).

Endpoint summary

Tips

• Upload is the prerequisite for transfer_method: "local_file". For files already on the web, skip upload and use transfer_method: "remote_url" with a url.
• Audio endpoints need the corresponding STT/TTS model enabled in the app's settings; otherwise you'll get a model/config error (see troubleshoot).