feat: NDI support & input sources abstraction by gioelecerati · Pull Request #439 · daydreamlive/scope

gioelecerati · 2026-02-11T08:51:33Z

Changes

Add NDI support with ctypes
- NDI sdk detection for darwin, linux, win
- Use native resolution
InputSource abstraction used by both the new NDI receiver and the existing spout receiver
OutputSink abstraction - spout sender moved into it
Frontend changes

How to test

Install the NDI sdk or NDI tools on your machine
Expose a NDI source (you can use for example OBS with the DIstroAV plugin enabling the "Preview" output, or install Resolume Arena and enable NDI output from the menu)
Start scope
Video input -> NDI -> Select the desired source
Play

In a follow-up PR

NDI sender for the OutputSink

Signed-off-by: gioelecerati <[email protected]>

leszko

I skimmed through the backend part. Nice work @gioelecerati I think it's in the good direction, I'll do the complete review when you finish your PR.

Some early comments:

I see that right now you copy a lot of logic from the Spout received logic, but do I understand correctly that you plan to replace the spout logic and now we'll only use the input source?
Now you've implemented the input, but I guess you're planning to do the same for the output. Is that correct?
As a side note: frame_processor.py is pretty overloaded. I see that you created a separate package core/inputs. That's great. Also when you work on _SpoutFrame class (or whatever you rename it to, consider moving it to a separate file. The same with some functions that are not strictly related to FrameProcessor. Thanks.

leszko · 2026-02-11T14:11:46Z

src/scope/server/frame_processor.py

+        # Handle generic input source settings
+        if "input_source" in parameters:
+            input_source_config = parameters.pop("input_source")
+            self._update_input_source(input_source_config)


Do we plan to allow changing the input_source when the stream is ongoing?

These parameters here are like "runtime" parameters. There are also the "intial_parameters" used during the initialization.

On the UI I am disallowing to switch from NDI to other inputs and viceversa, but it's just a patch because I am not sure how resolution changes are handled - for example, my NDI source was 1920x1080, switching to the 512x512 example video kept the output video as 1920x1080 but stretched, if you have any insight about this I will work to improve it

leszko · 2026-02-11T14:12:37Z

src/scope/server/frame_processor.py

+    return sources
+
+
 class _SpoutFrame:


If it's generic now, then I think we should rename it. Also, I think you can move it out of frame_processor.py

gioelecerati · 2026-02-11T16:00:29Z

I see that right now you copy a lot of logic from the Spout received logic, but do I understand correctly that you plan to replace the spout logic and now we'll only use the input source?

Yes, the spout receiver should use the new input source abstraction in the end

Now you've implemented the input, but I guess you're planning to do the same for the output. Is that correct?

Yes! Would like to make a follow-up PR making a NDI sender. I need to hook it up with the NDI sdk detection, so we can display the option only if the runtime is running on the machine

Signed-off-by: gioelecerati <[email protected]>

gioelecerati · 2026-02-12T06:42:58Z

@leszko Marked this as ready for review after addressing your comments and tested it on windows
Some notes:

I created a output sink and moved the spout sender there. The syphon sender is not showing up for me on my mac, is that known issue? I don't see it on the main branch as well
Removed the unused classes and methods, frame processor is back at roughly the size it had before these changes - let me know if you would break it down even more

leszko

Very nice PR @gioelecerati . Tested NDI and Spout in Windows and everything works smoothly! 🎉

Added some inline comments, but other than that the PR looks good.

frontend/src/components/InputAndControlsPanel.tsx

leszko · 2026-02-12T15:42:02Z

frontend/src/lib/api.ts

+export const getInputSourceResolution = async (
+  sourceType: string,
+  identifier: string,
+  timeoutMs = 5000
+): Promise<InputSourceResolution> => {
+  const response = await fetch(
+    `/api/v1/input-sources/${sourceType}/sources/${encodeURIComponent(identifier)}/resolution?timeout_ms=${timeoutMs}`,
+    {
+      method: "GET",
+      headers: { "Content-Type": "application/json" },
+    }
+  );
+
+  if (!response.ok) {
+    const errorText = await response.text();
+    throw new Error(
+      `Probe resolution failed: ${response.status} ${response.statusText}: ${errorText}`
+    );
+  }
+
+  return response.json();
+};


How did we handle the resolution before? I mean the resolution that the Spout sent, didn't we need a similar logic?

No Spout wasn't probing, it just had 512x512 hardcoded. Which I tried to do the same with NDI, but I was having the output trimmed or stretched depending on how I was applying it

This is also part of the reason why I went for not allowing source switch while you're on NDI, since the output will have the initial parameters resolution set as the NDI source resolution, ending up with a stretched output if you switch to camera or other source

Is there a way to change the resolution mid stream or are we planning to have that hot reloading parameters for w & h?

Either case, if it's possible to change res, I will address it in a follow up PR

src/scope/server/app.py

src/scope/server/frame_processor.py

leszko · 2026-02-12T15:56:33Z

src/scope/core/frames.py

+    """Wraps a raw numpy array to match the VideoFrame interface.
+
+    Allows code that expects a VideoFrame (with .to_ndarray()) to work
+    with plain numpy arrays from input sources like Spout and NDI.


I wonder for Spout, do we need the conversion from Tensor to VideoFrame? Isn't Spout all about keeping everything in GPU?

Ooof actually this rawframe is not used anymore, I will delete it

Not sure if I am knowledgeable enough about spout to answer your question tho, can you elaborate?

src/scope/server/app.py

src/scope/core/frames.py

src/scope/server/app.py

Signed-off-by: gioelecerati <[email protected]>

feat: NDI support & input sources abstraction

8b721a9

Signed-off-by: gioelecerati <[email protected]>

gioelecerati force-pushed the gio/feat/ndi branch from 860abd5 to 8b721a9 Compare February 11, 2026 08:55

gioelecerati added 3 commits February 11, 2026 09:59

prettier & remove string annotation

ceb93c8

Signed-off-by: gioelecerati <[email protected]>

ruff format

7a12e87

Signed-off-by: gioelecerati <[email protected]>

thumbnail preview & uiux fixes

66409e6

Signed-off-by: gioelecerati <[email protected]>

leszko reviewed Feb 11, 2026

View reviewed changes

output sink & remove spout frame

6dc3a26

Signed-off-by: gioelecerati <[email protected]>

gioelecerati marked this pull request as ready for review February 12, 2026 06:38

gioelecerati requested a review from leszko February 12, 2026 06:38

leszko reviewed Feb 12, 2026

View reviewed changes

gioelecerati added 2 commits February 12, 2026 17:31

address comments

4e0a69f

Signed-off-by: gioelecerati <[email protected]>

keep static image preview

69bcb6f

Signed-off-by: gioelecerati <[email protected]>

gioelecerati merged commit 8ab4261 into main Feb 12, 2026
6 checks passed

gioelecerati mentioned this pull request Feb 13, 2026

[experiment]: input sources plugins #434

Closed

Conversation

gioelecerati commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

leszko left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

leszko Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

gioelecerati Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

leszko Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

gioelecerati commented Feb 11, 2026

Uh oh!

gioelecerati commented Feb 12, 2026

Uh oh!

leszko left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

leszko Feb 12, 2026

Choose a reason for hiding this comment

Uh oh!

gioelecerati Feb 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

leszko Feb 12, 2026

Choose a reason for hiding this comment

Uh oh!

gioelecerati Feb 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

gioelecerati commented Feb 11, 2026 •

edited

Loading

leszko left a comment •

edited

Loading