String/Text Output response

[TRONBFFL]

I would like to be able to save text and have it in the message response as opposed to base64 for images as it is now. Just trying to use moondream to do image to text for me, but the runpod will only seem to work if an image is generated. Trying to figure out how to get the text to go back with the api reponse.

Many thanks

[esuvari]

src/rp_handler.py needs to be modified to handle other output formats beside image.

I have the same problem with video outputs. Trying to write my own video output handler.

Comfy API returns a JSON result such as the one below at the end of every job: {'9': {'images': [{'filename': 'ComfyUI_00010_.png', 'subfolder': '', 'type': 'output'}]}}

But it's different when the output is a video, the key becomes "gifs" instead of "images" {'10': {'gifs': [{'filename': 'AnimateDiff_00012.mp4', 'subfolder': '', 'type': 'output', 'format': 'video/h264-mp4', 'frame_rate': 24.0, 'workflow': 'AnimateDiff_00012.png', 'fullpath': '/ComfyUI/output/AnimateDiff_00012.mp4'}]}}

I imagine it'll be something else when the output is a .txt file.

I don't know what other keys are there but at the end of the day, it is rp_handler.py that needs to handle those outputs for what you're asking.

PS: While at it it'd be great to make a modification to accept various other input formats as well. (Audio etc.)

[TRONBFFL]

I finally got a version of rp handler to find the text - took me a bit though. For videos extensions I am saving to an azure blob at the end of my comfy workflow.

[AkshayLaghate]

I finally got a version of rp handler to find the text - took me a bit though. For videos extensions I am saving to an azure blob at the end of my comfy workflow.

Can you share the solution? I am having the same issue.

[TimPietrusky]

@TRONBFFL @esuvari @AkshayLaghate in the latest update, the way the output is detected and handled has been changed. i will take a look at your use case next, to make sure that "text" and "video" are also handled correctly.

@TRONBFFL would you mind to share the workflow you have screenshotted?

[TRONBFFL]

Sure here is a workflow I use.

moondream_wf.json

[vectris-dev]

Any progress here? I recently used this for images and it ran so well, now I have to do video outputs and I'd love to use it again.

[francoisoger]

Hello here,

I have the same problem, I need to get the result of a VHS_VideoCombine node, which is available in the gifs level of output.

So I'm curious about the update on this issue please