1
0
forked from cgvr/DeltaVR

update READMEs

This commit is contained in:
2026-05-13 14:14:28 +03:00
parent 7c67a4b44b
commit 34c693aaa3
4 changed files with 37 additions and 83 deletions
+16 -34
View File
@@ -1,38 +1,20 @@
### TODO
* pidev krabamise heli, kui kahe käega korraga krabada: "(Un)requesting ownership of Cube"
* FMOD ChannelControl errorid
* keyboard numbrid ei tööta pärast minigame'i completemist?
* speech-to-text:
* vahepeal lakkab töötamast lih??? tundub et siis kui pikalt tühjust salvestab? ei, vahepeal läheb kohe alguses ka
* sööta talle nulle kui on disabled, sest ikka kuidagi tuvastab eelnevat rääkimist
* uurida miks buildis Whisper halvemini töötab
* npc character:
* klaas on näha temast eespool
* shape detection:
* peab mängijale kuidagi selgitama, kuidas scale'ida prinditud objekte
* prinditud objekti scale'imisele min ja max size limiit
* 3d printerile soft particle'id, et ei oleks teravaid ääri seina sisse minnes
* archery range:
* võiks jääda kordama viimast instruktsiooni, kui mängija ei progressi edasi
* UFO möödalendamise animation/cutscene alguses kui ütleb "delta attacked by ufos" + panna juba alguses kapsli sisse ufo pöörlema
* cafe:
* võiks saada hinnata saadud tellimust
* "order again" nupp
# Testing the pipeline
Can't/Won't Do:
* glTF loading: vahetada ära shader Universal render pipeline Lit, mitte panna buildi kaasa glTf oma - **ei saa, objekt on siis ilma tekstuurita, lihtsalt hall**
Install required dependency libraries using
```
pip install -r requirements.txt
```
Make a copy of `.env.example`, rename it as `.env`, and insert the necessary config values.
### Notes
* Functionalities that were implemented using major help from AI:
* animating NPC mouth movement based on voice amplitude
* adjusting spawned gltf objects' scale
* porting InvokeAI client and TRELLIS client to Unity scripts
* Getting Whisper stream to work with FMOD instead of Unity default audio
* TRELLIS: added functionality to specify texture baking optimisation total steps as an argument (`texture_opt_total_steps`), to replace the hardcoded 2500. But this is not tracked in Git (because modified this https://github.com/IgorAherne/trellis-stable-projectorz/releases/tag/latest)
* Custom Shader Variant Collection to include glTF-pbrMetallicRoughness shader in build
## Python notebooks
Katse tähelepanekud:
* peab väga proper english accenti tegema
* mõne inimese puhul kuuleb "cat" asemel "cut"
* kui helitugevus oli liiga suur siis cafe waiter kuulis kohviku taustaheli taldriku klirinat kui "piano music"
* [`local_image_generation.ipynb`](notebooks/local_image_generation.ipynb)
* [`local_model_generation.ipynb`](notebooks/local_model_generation.ipynb)
* [`subprocess_test.ipynb`](notebooks/subprocess_test.ipynb)
## Python scripts
* [`generate_image_local.py`](generate_image_local.py)
* [`generate_model_local.py`](generate_model_local.py)
* [`start_pipeline.py`](start_pipeline.py)
-49
View File
@@ -1,49 +0,0 @@
3.170447962475009263e-05
2.923115389421582222e-04
2.510738931596279144e-02
1.111792679876089096e-02
6.692767888307571411e-02
1.006313711404800415e-01
7.780694961547851562e-02
5.686730891466140747e-02
5.614304915070533752e-02
4.554714635014533997e-02
4.514135792851448059e-02
5.479728057980537415e-02
4.272024706006050110e-02
3.989587724208831787e-02
4.298635944724082947e-02
4.074911773204803467e-02
2.244980260729789734e-02
1.105279754847288132e-02
1.347438804805278778e-02
1.654553040862083435e-02
1.846965588629245758e-02
2.045047841966152191e-02
1.407719496637582779e-02
6.578906439244747162e-03
1.353173051029443741e-02
1.625132374465465546e-02
5.863697826862335205e-02
1.110599413514137268e-01
9.950184077024459839e-02
1.184522062540054321e-01
1.000181213021278381e-01
6.772108376026153564e-02
7.621638476848602295e-02
3.018615581095218658e-02
9.624224901199340820e-02
1.259753555059432983e-01
1.276500672101974487e-01
1.206035763025283813e-01
1.011835709214210510e-01
6.155343726277351379e-02
3.734333068132400513e-02
2.485111355781555176e-02
2.122259326279163361e-02
1.139380130916833878e-02
7.472451310604810715e-03
5.807624198496341705e-03
1.960268709808588028e-03
8.761089411564171314e-04
3.071058890782296658e-04
+21
View File
@@ -43,6 +43,27 @@ In both editor mode and the built version, after first starting the game, the `c
* `invokeAiModelKey` - the key of the specific model set up in InvokeAI to use. To see the keys of your installed models, send a GET request to `{INVOKEAI_URL}/api/v2/models/?model_type=main` (for example, `http://192.168.0.53:9090/api/v2/models/?model_type=main`)
* `trellisUrl` - the URL of TRELLIS (for example, `http://192.168.0.53:7960`)
The [3d-generation-pipeline](3d-generation-pipeline/README.md) folder includes Python notebooks and scripts that can be used to test the pipeline working together.
## Unresolved bugs and potential improvements for the future
* continuous grabbing sound when grabbing with both hands at the same time: "(Un)requesting ownership of Cube"
* FMOD ChannelControl errors
* speech-to-text: investigate why Whisper still sometimes stops detecting speech
* glTF loading: investigate how to use the Universal Render Pipeline Lit shader for the glTF-loaded objects.
* NPC characters:
* glass wall is sometimes incorrectly visible in front of the NPC
* Delta Cafe:
* ability for players to rate the orders
* "order again" button
* Archery Range:
* NPC should repeat the last instruction when the player is not progressing
* reuse the UFO flying animation from Space Walk in the beginning, when NPC talks about looking out the window and the UFO attack + have the default UFO already in the model display capsule
* Shape Detection:
* explain to the player how to scale printed objects
* set min and max scale limits on printed objects
* set 3d printer soft particles to avoid sharp edges when particles go inside the printer wall
## Credits
**Samuel Saar**<br/>