Video Generation failing

[copypastegod]

Updated to newest Nvidia Studio Drivers 595.79

2026-04-03 06:24:45,189 - INFO - [Electron] Session log file: C:\Users\Admin\AppData\Local\LTXDesktop\logs\session_2026-04-02_17-24-45_unknown.log
2026-04-03 06:24:45,227 - INFO - [Electron] [icon] Loading app icon from: C:\AI\LTX\LTX Desktop\resources\icon.ico | exists: false
2026-04-03 06:24:45,323 - INFO - [Renderer] Projects saved: 1
2026-04-03 06:24:45,324 - INFO - [Renderer] Starting Python backend...
2026-04-03 06:24:45,325 - INFO - [Electron] Using bundled Python: C:\Users\Admin\AppData\Local\LTXDesktop\python\python.exe
2026-04-03 06:24:45,325 - INFO - [Electron] Starting Python backend: C:\Users\Admin\AppData\Local\LTXDesktop\python\python.exe C:\AI\LTX\LTX Desktop\resources\backend\ltx2_server.py
2026-04-03 06:24:45,786 - ERROR - [Backend] C:\Users\Admin\AppData\Local\LTXDesktop\python\Lib\site-packages\torch\cuda\__init__.py:65: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
2026-04-03 06:24:45,786 - ERROR - [Backend]   import pynvml  # type: ignore[import]
2026-04-03 06:24:46,656 - INFO - [Backend] INFO:__main__:SageAttention enabled - attention operations will be faster
2026-04-03 06:24:46,716 - INFO - [Backend] INFO:__main__:Models directory: C:\Users\Admin\AppData\Local\LTXDesktop\models
2026-04-03 06:24:46,745 - INFO - [Backend] INFO:__main__:Runtime policy force_api_generations=False (system=Windows cuda_available=True vram_gb=15)
2026-04-03 06:24:49,169 - INFO - [Backend] INFO:handlers.settings_handler:Settings loaded from C:\Users\Admin\AppData\Local\LTXDesktop\settings.json
2026-04-03 06:24:49,189 - INFO - [Backend] INFO:__main__:============================================================
2026-04-03 06:24:49,190 - INFO - [Backend] INFO:__main__:LTX-2 Video Generation Server (FastAPI + Uvicorn)
2026-04-03 06:24:49,211 - INFO - [Backend] INFO:__main__:Platform: Windows (AMD64)
2026-04-03 06:24:49,211 - INFO - [Backend] INFO:__main__:Device: cuda  |  Dtype: torch.bfloat16
2026-04-03 06:24:49,211 - INFO - [Backend] INFO:__main__:GPU: NVIDIA GeForce RTX 5080  |  VRAM: 15 GB
2026-04-03 06:24:49,211 - INFO - [Backend] INFO:__main__:SageAttention: enabled
2026-04-03 06:24:49,211 - INFO - [Backend] INFO:__main__:Python: 3.13.12  |  Torch: 2.10.0+cu128
2026-04-03 06:24:49,212 - INFO - [Backend] INFO:__main__:============================================================
2026-04-03 06:24:49,231 - INFO - [Backend] Started server process [17040]
2026-04-03 06:24:49,231 - INFO - [Backend] Waiting for application startup.
2026-04-03 06:24:49,231 - INFO - [Backend] Application startup complete.
2026-04-03 06:24:49,232 - INFO - [Backend] Server running on http://127.0.0.1:14300
2026-04-03 06:24:49,232 - INFO - [Renderer] Checking backend health...
2026-04-03 06:24:49,233 - INFO - [Renderer] Python backend started successfully
2026-04-03 06:24:49,258 - INFO - [Renderer] Backend health: {"status":"ok","models_loaded":false,"active_model":null,"gpu_info":{"name":"NVIDIA GeForce RTX 5080","vram":16303,"vramUsed":726},"sage_attention":true,"models_status":[{"id":"fast","name":"LTX-2 Fast (Distilled)","loaded":false,"downloaded":true}]}
2026-04-03 06:24:49,400 - INFO - [Backend] INFO:_routes.settings:Applied settings patch (changed=none)
2026-04-03 06:24:50,255 - INFO - [Electron] Checking for update...
2026-04-03 06:24:52,320 - INFO - [Backend] INFO:handlers.video_generation_handler:Resolution 540p - using fast pipeline
2026-04-03 06:24:52,376 - INFO - [Backend] INFO:handlers.video_generation_handler:Image: C:\Users\Admin\Downloads\Ltx Desktop Assets\project-1775148378860-pjzh4gnil\zit_image_20260403_061057_9a2049cc.png -> 960x544
2026-04-03 06:24:52,377 - INFO - [Backend] INFO:services.text_encoder.ltx_text_encoder:Installed PromptEncoder.__init__ patch for None gemma_root
2026-04-03 06:24:52,377 - INFO - [Backend] INFO:services.text_encoder.ltx_text_encoder:Installed PromptEncoder API embeddings patch
2026-04-03 06:24:52,377 - INFO - [Backend] WARNING:services.text_encoder.ltx_text_encoder:Failed to patch cleanup_memory for module ltx_pipelines.retake_pipeline
2026-04-03 06:24:52,378 - INFO - [Backend] Traceback (most recent call last):
2026-04-03 06:24:52,378 - INFO - [Backend]   File "C:\AI\LTX\LTX Desktop\resources\backend\services\text_encoder\ltx_text_encoder.py", line 156, in _install_cleanup_memory_patch
2026-04-03 06:24:52,378 - INFO - [Backend]     module = __import__(module_name, fromlist=["cleanup_memory"])
2026-04-03 06:24:52,378 - INFO - [Backend] ModuleNotFoundError: No module named 'ltx_pipelines.retake_pipeline'
2026-04-03 06:24:52,378 - INFO - [Backend] INFO:services.text_encoder.ltx_text_encoder:Installed cleanup_memory patch
2026-04-03 06:24:52,379 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Generation started (model=fast, 960x544, 121 frames, 24 fps)
2026-04-03 06:24:52,379 - INFO - [Backend] INFO:services.text_encoder.ltx_text_encoder:Installed PromptEncoder.__init__ patch for None gemma_root
2026-04-03 06:24:52,379 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Pipeline load: 0.00s
2026-04-03 06:24:52,422 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Text encoding (local): 0.00s
2026-04-03 06:24:53,352 - ERROR - [Backend] Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.52, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
2026-04-03 06:25:50,109 - ERROR - [Backend] 
  0%|          | 0/8 [00:00<?, ?it/s]
2026-04-03 06:26:04,768 - ERROR - [Backend] 
 12%|#2        | 1/8 [00:14<01:42, 14.67s/it]
2026-04-03 06:26:06,095 - ERROR - [Backend] 
 25%|##5       | 2/8 [00:16<00:40,  6.82s/it]
2026-04-03 06:26:07,422 - ERROR - [Backend] 
 38%|###7      | 3/8 [00:17<00:21,  4.31s/it]
2026-04-03 06:26:08,772 - ERROR - [Backend] 
 50%|#####     | 4/8 [00:18<00:12,  3.14s/it]
2026-04-03 06:26:10,096 - ERROR - [Backend] 
 62%|######2   | 5/8 [00:20<00:07,  2.49s/it]
2026-04-03 06:26:11,420 - ERROR - [Backend] 
 75%|#######5  | 6/8 [00:21<00:04,  2.09s/it]
2026-04-03 06:26:12,746 - ERROR - [Backend] 
 88%|########7 | 7/8 [00:22<00:01,  1.84s/it]
2026-04-03 06:26:14,069 - ERROR - [Backend] 
100%|##########| 8/8 [00:23<00:00,  1.68s/it]
2026-04-03 06:26:14,070 - ERROR - [Backend] 
100%|##########| 8/8 [00:23<00:00,  3.00s/it]
2026-04-03 06:26:17,964 - INFO - [Electron] Python backend exited with code 3221225477

[ofirgluzman]

copypastegod is the issue consistently reproduced?

[jtrogo]

Exact same error on mine. NVidia driver 595.97

RTX 5080 + 15 GB detected VRAM
Torch 2.10.0+cu128 on bundled Python 3.13
SageAttention enabled

Windows access violation 0xC0000005 in torch_cpu.dll

session_2026-04-02_17-25-08_unknown.log

[copypastegod]

Yes, it fails everytime.

It seems to generate the video but fails towards the end. Reinstalled twice, tried turning Hardware Accelerated GPU Scheduler on, using API encoder etc... but still the same error.

2026-04-03 07:26:19,648 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Generation started (model=fast, 960x544, 121 frames, 24 fps)
2026-04-03 07:26:19,648 - INFO - [Backend] INFO:services.text_encoder.ltx_text_encoder:Installed PromptEncoder.__init__ patch for None gemma_root
2026-04-03 07:26:19,648 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Pipeline load: 0.00s
2026-04-03 07:26:19,696 - INFO - [Backend] INFO:handlers.video_generation_handler:[i2v] Text encoding (local): 0.00s
2026-04-03 07:26:20,610 - ERROR - [Backend] Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.52, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
2026-04-03 07:27:12,189 - ERROR - [Backend] 
  0%|          | 0/8 [00:00<?, ?it/s]
2026-04-03 07:27:26,037 - ERROR - [Backend] 
 12%|#2        | 1/8 [00:13<01:36, 13.85s/it]
2026-04-03 07:27:27,353 - ERROR - [Backend] 
 25%|##5       | 2/8 [00:15<00:38,  6.48s/it]
2026-04-03 07:27:28,665 - ERROR - [Backend] 
 38%|###7      | 3/8 [00:16<00:20,  4.12s/it]
2026-04-03 07:27:29,984 - ERROR - [Backend] 
 50%|#####     | 4/8 [00:17<00:12,  3.01s/it]
2026-04-03 07:27:31,302 - ERROR - [Backend] 
 62%|######2   | 5/8 [00:19<00:07,  2.40s/it]
2026-04-03 07:27:32,619 - ERROR - [Backend] 
 75%|#######5  | 6/8 [00:20<00:04,  2.03s/it]
2026-04-03 07:27:33,942 - ERROR - [Backend] 
 88%|########7 | 7/8 [00:21<00:01,  1.80s/it]
2026-04-03 07:27:35,261 - ERROR - [Backend] 
100%|##########| 8/8 [00:23<00:00,  1.65s/it]
2026-04-03 07:27:35,261 - ERROR - [Backend] 
100%|##########| 8/8 [00:23<00:00,  2.88s/it]
2026-04-03 07:27:38,203 - INFO - [Electron] Python backend exited with code 3221225477

Image generation working fine however... only Video generation crashes the backend.

[Ionaut]

Same Errors for me as well!
RTX 3090 24GB
64 GB RAM
Win11

[SnoopGER]

jup after a few good tries backend stopped working even after restart

[Aerasyn]

Same exact error. nvidia 5080. driver 595.97

[Teutonick]

same here 4070 16gb. driver 595.97

[vegevil12314]

Having the exact same issue with a 5090 driver 595.97

[ofirgluzman]

Hi jtrogo, copypastegod, Ionaut (@Ionaut), SnoopGER, Aerasyn, mr.andr (@Teutonick), vegevil12314 — thanks for reporting this issue.

I believe v1.0.4 should resolve the backend crash during video generation. Could you please download and install the latest version and let me know if the issue persists?

Download link: https://github.com/Lightricks/LTX-Desktop/releases/download/v1.0.4/LTX-Desktop-Setup.exe

Please report back here whether it fixes the problem for you. Thanks!

[copypastegod]

Thanks, seems to work now off initial testing!

[jtrogo]

It's working for me as well. I2V, 20s , 540, 16:9. Have not tested other configurations.

Thank you for the quick fix!

[vegevil12314]

Works after the update on my 5090. Thank you.

[SnoopGER]

Hi jtrogo, copypastegod, Ionaut (@Ionaut), SnoopGER, Aerasyn, mr.andr (@Teutonick), vegevil12314 — thanks for reporting this issue.

I believe v1.0.4 should resolve the backend crash during video generation. Could you please download and install the latest version and let me know if the issue persists?

Download link: https://github.com/Lightricks/LTX-Desktop/releases/download/v1.0.4/LTX-Desktop-Setup.exe

Please report back here whether it fixes the problem for you. Thanks!

Thank you, its fixed (5090)

[Teutonick]

now it is OK!!! THANX)))