Animating scene audio volume doesn't work #75686

New Issue

ChristopherAnderssarian · 2020-04-13T17:55:22+02:00

2020-04-13 17:55:22 +02:00

Title: Animating scene audio volume dosn't work
System Information
Operating system: Windows 7
GPU: Radeon Vega 64, Divider version: 19.12.2
CPU: Intel Core i7-5960X

Blender Version(s)
(listed are versions tested)

Working: 2.79 (sub 00), branch: master, commit date: 2018-03-22 14:10, hash: f4dc9f9d68b, type: Release, build date: 2018-03-22, 09:59 AM
Broken: 2.80 (sub 75), branch: master, commit date: 2019-07-29 14:47, hash: f6cb5f54494e, type: Release, build date: 2019-07-29, 09:44 AM
Broken: 2.81 (sub 16), branch: master, commit date: 2019-11-20 14:27, hash: 26bd5ebd42e3, type: Release, build date: 2019-11-20, 16:33:00
Broken: 2.82 (sub 07), branch: master, commit date: 2020-02-12 16:20, hash: 77d23b0bd76f, type: Release, build date: 2020-02-13, 01:56:46
Broken: blender-2.83-2e75172c45fd-windows64 (build bot)

Short description of error
Animating scene audio volume doesn't affect the preview or render.

Exact steps for others to reproduce the error

Scene_Audio_keying.blend

Title: Animating scene audio volume dosn't work **System Information** Operating system: Windows 7 GPU: Radeon Vega 64, Divider version: 19.12.2 CPU: Intel Core i7-5960X **Blender Version(s)** *(listed are versions tested)* Working: `2.79 (sub 00), branch: master, commit date: 2018-03-22 14:10, hash: f4dc9f9d68b, type: Release, build date: 2018-03-22, 09:59 AM` Broken: ` 2.80 (sub 75), branch: master, commit date: 2019-07-29 14:47, hash: f6cb5f54494e, type: Release, build date: 2019-07-29, 09:44 AM` Broken: ` 2.81 (sub 16), branch: master, commit date: 2019-11-20 14:27, hash: 26bd5ebd42e3, type: Release, build date: 2019-11-20, 16:33:00` Broken: ` 2.82 (sub 07), branch: master, commit date: 2020-02-12 16:20, hash: 77d23b0bd76f, type: Release, build date: 2020-02-13, 01:56:46` Broken: `blender-2.83-2e75172c45fd-windows64` (build bot) **Short description of error** Animating scene audio volume doesn't affect the preview or render. **Exact steps for others to reproduce the error** [Scene_Audio_keying.blend](https://archive.blender.org/developer/F8469287/Scene_Audio_keying.blend)

ChristopherAnderssarian commented

2020-04-13 17:55:22 +02:00

Added subscriber: @ChristopherAnderssarian

Philipp Oeser commented

2020-04-14 12:21:35 +02:00

Added subscriber: @lichtwerk

Philipp Oeser commented

2020-04-14 12:21:35 +02:00

Changed status from 'Needs Triage' to: 'Confirmed'

Philipp Oeser commented

2020-04-14 12:21:35 +02:00

Can confirm.
Manually editing the scene audio volume works, doing this with keyframes doesnt.

Not quite sure why this is set to Low priority? Would actually set this to High instead (seems quite important, no?)...

Can confirm. Manually editing the scene audio volume works, doing this with keyframes doesnt. Not quite sure why this is set to Low priority? Would actually set this to High instead (seems quite important, no?)...

ChristopherAnderssarian commented

2020-04-14 12:44:27 +02:00

I used the wrong phab link, apologies. With the amount of reports I've submitted being deemed low priority I assumed this to be a "obscure workflow", but having said that; thanks for the triage!

Sybren A. Stüvel self-assigned this 2020-04-14 16:12:51 +02:00

Philipp Oeser commented

2020-04-14 16:21:47 +02:00

Added subscriber: @Sergey

Philipp Oeser commented

2020-04-14 16:21:47 +02:00

After all this time, I am still confused about our animation system and dependency graph...

So Animation seems to be correctly executed:

BKE_animsys_eval_animdata
BKE_animsys_write_rna_setting
RNA_property_float_set
fprop->set

all seem to do their thing...

For the property_update (rna_Scene_volume_update) , this is set up to tag the scene with ID_RECALC_AUDIO_VOLUME, however in BKE_scene_update_sound [which is constantly called] this recalc flag doesnt seem to be set on the evaluated scene... so BKE_sound_set_scene_volume isnt called...

I feel dumb, why is this so? Maybe @dr.sybren or @Sergey would know...

After all this time, I am still confused about our animation system and dependency graph... So Animation seems to be correctly executed: - BKE_animsys_eval_animdata - BKE_animsys_write_rna_setting - RNA_property_float_set - fprop->set all seem to do their thing... For the property_update (`rna_Scene_volume_update`) , this is set up to tag the scene with `ID_RECALC_AUDIO_VOLUME`, however in `BKE_scene_update_sound` [which is constantly called] this recalc flag doesnt seem to be set on the evaluated scene... so `BKE_sound_set_scene_volume` isnt called... I feel dumb, why is this so? Maybe @dr.sybren or @Sergey would know...

Philipp Oeser commented

2020-04-14 18:49:14 +02:00

On the sideline:
Was wondering if this should be removed?
6a392e8cb5

Was disabled / removed in
b062056c05, 21744217ce

On the sideline: Was wondering if this should be removed? 6a392e8cb5 Was disabled / removed in b062056c05, 21744217ce

blender-admin commented

2020-04-14 19:22:18 +02:00

This issue was referenced by 6adb254bb0

This issue was referenced by 6adb254bb046ab337cfd4225ca55e17f196db312

Sergey Sharybin commented

2020-04-15 09:14:10 +02:00

@lichtwerk, noit sure what exactly you mean. The mechanism where graph evaluation tags graph for updates (aka animation system using update callback) is very weak and fragile.

For the original report what we can do is to introduce dedicated audio properties which will assign recalc flags on the evaluated scene, similar to how it's done for individual particle systems which then gets handled by a more "uber" update function.

@lichtwerk, noit sure what exactly you mean. The mechanism where graph evaluation tags graph for updates (aka animation system using update callback) is very weak and fragile. For the original report what we can do is to introduce dedicated audio properties which will assign recalc flags on the evaluated scene, similar to how it's done for individual particle systems which then gets handled by a more "uber" update function.

Philipp Oeser commented

2020-04-15 10:31:43 +02:00

In #75686#910185, @Sergey wrote:
@lichtwerk, noit sure what exactly you mean. The mechanism where graph evaluation tags graph for updates (aka animation system using update callback) is very weak and fragile.

I meant removing all traces to the old code, see D7432: Cleanup: remove (unused) RNA update cache

> In #75686#910185, @Sergey wrote: > @lichtwerk, noit sure what exactly you mean. The mechanism where graph evaluation tags graph for updates (aka animation system using update callback) is very weak and fragile. I meant removing all traces to the old code, see [D7432: Cleanup: remove (unused) RNA update cache](https://archive.blender.org/developer/D7432)

Sybren A. Stüvel commented

2020-04-17 10:24:02 +02:00

In #75686#910185, @Sergey wrote:
For the original report what we can do is to introduce dedicated audio properties which will assign recalc flags on the evaluated scene, similar to how it's done for individual particle systems which then gets handled by a more "uber" update function.

How would you design such dedicated audio properties? How would it differ from the tagging that's done in D7429 in rna_Scene_use_audio_set()?

> In #75686#910185, @Sergey wrote: > For the original report what we can do is to introduce dedicated audio properties which will assign recalc flags on the evaluated scene, similar to how it's done for individual particle systems which then gets handled by a more "uber" update function. How would you design such dedicated audio properties? How would it differ from the tagging that's done in [D7429](https://archive.blender.org/developer/D7429) in `rna_Scene_use_audio_set()`?

Sergey Sharybin commented

2020-04-17 15:08:50 +02:00

How would you design such dedicated audio properties?

Add OperationCode::AUDIO_VOLUME operation to NodeType::AUDIO component with the callback which is explained in a bit.
Add relation from OperationCode::AUDIO_VOLUME to OperationCode::SOUND_EVAL.
Add relation from animation to OperationCode::AUDIO_VOLUME if the volume is animated

The callback for OperationCode::AUDIO_VOLUME would set a flag in the actual scene which needs this. It can set the same ID_RECALC_AUDIO_VOLUME.

Currently the sound is updated in BKE_scene_update_sound from either scene_graph_update_tagged or from BKE_scene_graph_update_for_newframe. This means, that doing that simple edit should be enough to get things going. In a longer perspective the BKE_scene_update_sound should be done as OperationCode::SOUND_EVAL callback, and also incorporate sequencer sound update.

How would it differ from the tagging that's done in D7429 in rna_Scene_use_audio_set()

It will differ in a sense that it will not violate the law [insert the cut from Kung Fury where he tells Cobra he is under arrest for breaking the law].

The RNA is designed in a way that set() is only responsible for modifying value in DNA. Tagging should be happening from update().
Doing dependency graph tagging during evaluation is meaningless. It might work in this simple case and at the current state of code. It will break when sound evaluation is properly moved as a dependency graph evaluation operation.
Dependency graph tag notifies all dependency graphs about the change, and you obviously don't want this to happen from inside some random dependency graph evaluation.

> How would you design such dedicated audio properties? 1. Add `OperationCode::AUDIO_VOLUME` operation to `NodeType::AUDIO` component with the callback which is explained in a bit. 2. Add relation from `OperationCode::AUDIO_VOLUME` to `OperationCode::SOUND_EVAL`. 3. Add relation from animation to `OperationCode::AUDIO_VOLUME` if the volume is animated The callback for `OperationCode::AUDIO_VOLUME` would set a flag in the actual scene which needs this. It can set the same `ID_RECALC_AUDIO_VOLUME`. Currently the sound is updated in `BKE_scene_update_sound` from either `scene_graph_update_tagged` or from `BKE_scene_graph_update_for_newframe`. This means, that doing that simple edit should be enough to get things going. In a longer perspective the `BKE_scene_update_sound` should be done as `OperationCode::SOUND_EVAL` callback, and also incorporate sequencer sound update. > How would it differ from the tagging that's done in [D7429](https://archive.blender.org/developer/D7429) in `rna_Scene_use_audio_set()` It will differ in a sense that it will not violate the law *[insert the cut from Kung Fury where he tells Cobra he is under arrest for breaking the law]*. - The RNA is designed in a way that `set()` is only responsible for modifying value in DNA. Tagging should be happening from `update()`. - Doing dependency graph tagging during evaluation is meaningless. It might work in this simple case and at the current state of code. It will break when sound evaluation is properly moved as a dependency graph evaluation operation. - Dependency graph tag notifies all dependency graphs about the change, and you obviously don't want this to happen from inside some random dependency graph evaluation.

Sybren A. Stüvel commented

2020-04-17 16:39:10 +02:00

Changed status from 'Confirmed' to: 'Resolved'

Sybren A. Stüvel closed this issue

2020-04-17 16:39:10 +02:00

Thomas Dinges added this to the 2.83 LTS milestone 2023-02-08 16:38:05 +01:00

Philipp Oeser referenced this issue

2024-04-19 16:18:42 +02:00

Regression: animated scene audio volume not changing the audible volume on playback #120726

Philipp Oeser referenced this issue

2024-04-19 16:27:00 +02:00

Regression: animated scene audio volume not changing the audible volume on playback #120726