Wan2GP fails to run after the latest v13.0 update

[NiceDragon8066]

I pulled in the latest update to the app going from stable v12.72 to the latest v13.0, The app fails to initialize with error messages. Please see below the error message I am getting in the system console at the time of launch. I would request instructions to fix this issue or please specify a method to roll back the changes to v12.72, which was working fine. Thank you.

--- Initializing Wan2GP on RTX 5090 (GPU 0) ---
Starting Wan2GP...
[GGUF][llama.cpp CUDA] kernels available.
[PluginManager] Loaded model plugin: Krea 2 Identity Edit (from wan2gp-krea2-identity-edit)
[Quanto][INT8] Injected int8 kernels ACTIVE (backend=triton).
Loaded plugin: Image Suite (from ImageSuite-Wan2GP)
Loaded plugin: LTX Director (from LTX-Director)

Prompt Library: patient localhost probe installed (prevents 'When localhost is not accessible' on large UIs) <<<
Loaded plugin: Prompt Library (from MediaGen-PromptLib-Wan2GP)
[H3RefMod] declared h3_refmod_state / h3_refmod_extract custom settings on MiniMax H3's model definition
[H3RefMod] patched validate_generative_settings so RefMods count as visual references against MiniMax H3's audio-reference pre-flight check
[H3RefMod] patches installed on MiniMaxH3Pipeline (RefMod extraction + injection enabled)
Loaded plugin: MiniMax H3 RefMods (from MiniMaxH3Mod-for-WanGP)
Loaded plugin: MiniMax H3 Prompt Builder (from Wan2GP-Minimax-H3-Prompt-Builder)
Loaded plugin: Prompt Builder (from Wan2GP-Prompt-Builder)
[FM] v3.6.4 Token:True
Loaded plugin: Finetune Manager (from Wan2GP-finetune-manager)
Loaded plugin: Media Flow (from media_flow)
Loaded plugin: Models Manager (from models_manager)
Loaded plugin: Motion Designer (from motion_designer)
Loaded plugin: File Gallery (from wan2gp-gallery)
Loaded plugin: Krea 2 Identity Edit (from wan2gp-krea2-identity-edit)
Loaded plugin: LoRA Manager (from wan2gp-lora-manager)
Loaded plugin: LoRA Merger (from wan2gp-lora-merger)
Loaded plugin: Lora Multipliers UI (from wan2gp-lora-multipliers-ui)
Loaded plugin: LoRA Organizer (from wan2gp-lora-organizer)
Loaded plugin: Multi-Angle Prompt Helper (from wan2gp-multiangle-prompt-helper)
Loaded plugin: Prompt Manager (from wan2gp-prompt-manager)
Loaded plugin: MiniMax Ref2VA Prompt Studio (from wan2gp-prompt-studio)
Loaded plugin: Wildcards (from wan2gp-wildcards)
[H3RefMod] refreshed Wan2GP's model catalog so the RefMod custom_settings declaration takes effect on cached MiniMax H3 model definitions
[H3RefMod] hooked prepare_inputs_dict so the inline RefMods panel on the Media Generator page persists across form edits
[H3RefMod] hooked get_model_settings so the inline panel's RefMod selection is always fresh at the moment Generate actually queues the task

ImageSuite UI build ORPHANSUITE-5 loaded (LoRAs in enhancements + per-family Default Generation Values) <<<

ImageSuite: relaxed Gradio file-cache check (existing local files under the allow-list now allowed) <<<
Traceback (most recent call last):
File "C:\Users\vidz3\Wan2GP\wgp.py", line 14177, in
demo = create_ui()
^^^^^^^^^^^
File "C:\Users\vidz3\Wan2GP\wgp.py", line 13773, in create_ui
with ui_studio.Blocks(css=css, js=js, theme=theme, title="WanGP", fill_width=True) as main:
File "C:\Users\vidz3.conda\envs\wan2gp\Lib\site-packages\gradio\blocks.py", line 2332, in exit
self.config = self.get_config_file()
^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\vidz3\Wan2GP\shared\gradio\ui_studio.py", line 111, in get_config_file
return style_config(super().get_config_file(), self.theme)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\vidz3\Wan2GP\shared\gradio\gradio_model_change_queue.py", line 62, in get_config_file
_prepare(self)
File "C:\Users\vidz3\Wan2GP\shared\gradio\gradio_model_change_queue.py", line 43, in _prepare
[tail] = [fn for fn in app.fns.values() if (target._id, 'change') in fn.targets and fn._id in parents]
^^^^^^
ValueError: too many values to unpack (expected 1)
Autosave: Queue is empty, nothing to save.

Wan2GP session ended. Press any key to close...

[TinyMikeG]

I had the same issue last night. I couldn't solve it so I reinstalled from scratch. I had backups of my chkpt folder and loras. The built-in reinstall didn't help, I had to completely uninstall and delete the install folder, download the desktop install file and start again. It solved it but there are issues with this version I've found. There is no progress update when generating it generated but says pending throughout. I also find it slightly slower. If there is an easier way someone might tell you, but this solved it.

[NiceDragon8066]

I had to reinstall as well. But now whole bunch of new and strange problems are happening. This new app treats Qwen-Image-Edit-2511 as a video model and is telling me that it will generate a minimum of 5 frames! Also, every T2I prompt encodes twice and the text encoder phase takes too long. The transformer denoising seems to be faster for some reason. Also, most of the plugins don't work anymore. Some of them cause app crash at startup with gradio error and even the models manager plugin fails with error. I hope deepbeepmeep knows what's wrong and fixes it soon. Sad really, a fine working app is now broken because of over-zealous unnecessary upgrade.

…

On Mon, 14 Sept 2026, 12:30 pm TinyMikeG, ***@***.***> wrote: *TinyMikeG* left a comment (deepbeepmeep/Wan2GP#2304) <#2304 (comment)> I had the same issue last night. I couldn't solve it so I reinstalled from scratch. I had backups of my chkpt folder and loras. The built-in reinstall didn't help, I had to completely uninstall and delete the install folder, download the desktop install file and start again. It solved it but there are issues with this version I've found. There is no progress update when generating it generated but says pending throughout. I also find it slightly slower. If there is an easier way someone might tell you, but this solved it. — Reply to this email directly, view it on GitHub <#2304?email_source=notifications&email_token=BYUOHLJXSPYDLRTACDPY2VD5O6JSNA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGAZDEOJVGYY2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-5660229561>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BYUOHLNDBOULK5IVW3UHST35O6JSNAVCNFSNUABFKJSXA33TNF2G64TZHM4TIMBSHE3DQMJVHNEXG43VMU5TKNBUGQYDIMJYG422C5QC> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/BYUOHLIKG4M64FEIZSJMIXT5O6JSNA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGAZDEOJVGYY2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/BYUOHLI5JHYFQYYK7W5TYID5O6JSNA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGAZDEOJVGYY2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[omeganebula666-creator]

the bug is caused by one of your plugins, that is how i solved it by editing the "wgp_config.json" file and removing from the enabled_plugins list each plugin one at a time until i found the culprit which in my case was "wan2gp-prompt-studio"

[NiceDragon8066]

Yeah, I did the same thing and it solved the problem. Thanks. But, I still don't know why the Qwen text encoder runs twice (shows Encoding 1/2, Encoding 2/2), even though there is just one single prompt without any line break. Also, still getting error when accessing the Models Manager plugin. Wasted half of my day reinstalling the app and struggling through all these newly incorporated bugs!

…

On Mon, 14 Sept 2026, 2:34 pm omeganebula666-creator, < ***@***.***> wrote: *omeganebula666-creator* left a comment (deepbeepmeep/Wan2GP#2304) <#2304 (comment)> the bug is caused by one of your plugins, that is how i solved it by editing the "wgp_config.json" file and removing from the enabled_plugins list each plugin one at a time until i found the culprit which in my case was "wan2gp-prompt-studio" — Reply to this email directly, view it on GitHub <#2304?email_source=notifications&email_token=BYUOHLJRFIFGOPRVJULK7ID5O6YBPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGE2TMNJSHA32M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-5661565287>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BYUOHLLSXDXCMYCSYWW2RFT5O6YBPAVCNFSNUABFKJSXA33TNF2G64TZHM4TIMBSHE3DQMJVHNEXG43VMU5TKNBUGQYDIMJYG422C5QC> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/BYUOHLK4BIYQ7X4YP4JAV6T5O6YBPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGE2TMNJSHA32M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/BYUOHLLFXSYUJT62JZKM3KD5O6YBPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRWGE2TMNJSHA32M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[Tophness]

File "C:\Users\vidz3.conda\envs\wan2gp\Lib\site-packages\gradio\blocks.py", line 2332, in exit

When you see errors like this, it's generally a plugin issue. I try to tell people to avoid gradio blocks when possible because they're very sensitive to gradio versions or minor changes in gradio, but the AI loves to use them.
Unfortunately it's one where safe mode can't prevent it yet, so you'll have to go in to your wgp_config.json file and remove the item from enabled_plugins that caused it. In this case it looks like LTX Director.

It appears V13 broke a lot of plugins. I just fixed mine.
It shouldn't be too hard to update them, but it's going to be a pain point for some until they do.

[Tophness]

which in my case was "wan2gp-prompt-studio"

When did this happen?
I've just recently updated prompt studio for V13, so you might want to try it again, but it didn't error out like that before for me.