We have identified a critical issue in SeaArt image generation where prompts are not reflected in images under mobile network (4G/5G) conditions.
【Observed Phenomenon】
・When generating images using mobile data on devices like iPhones, even if specific instructions like "Ukiyo-e style, Hannya mask, dragon tattoo" are entered in the prompt and these strings are correctly recorded in the post-generation properties, the output image completely ignores them and produces a default-like image (e.g., a simple close-up of a person).
・Using the same prompt, settings, and model, the image generates correctly when using a WiFi connection.
【Issue】
The system forces generation without reporting errors, consuming resources (points or free attempts), even when it fails to correctly receive and process prompt data.
There is a clear discrepancy between the data in the properties (logs) and the actual generation results, indicating a breakdown in system consistency.
【Request】
This issue appears to stem from network-related data transmission failures or insufficient server-side validation. As it causes significant disadvantage to general users consuming points, we strongly demand immediate investigation and the implementation of a process to halt generation when incomplete data is received.


Please authenticate to join the conversation.
Need Customer Service
Feedback
Works/Post/Article
24 days ago

七波
Get notified by email when there are changes.
Need Customer Service
Feedback
Works/Post/Article
24 days ago

七波
Get notified by email when there are changes.