Google Flow / Veo 3.1 – Prompt Forgetting Issue (Context Loss, Not Following Instructions)

Bug Report: Google Flow / Veo 3.1 – Prompt Ignoring and Context Loss Issue

I am reporting a recurring issue in Google Flow / Veo 3.1 where the system fails to reliably follow the given prompt and loses context during generation.

Despite providing detailed and structured instructions (including scene composition, character positioning, outfits, actions, and frame constraints), the model frequently deviates from the original input. It appears to partially ignore or overwrite specific instructions, resulting in outputs that do not match the requested frame or scenario.

In addition, the model does not consistently maintain continuity across the prompt. Even when the instructions are explicit, the generated result may introduce unintended changes in behavior, framing, or visual consistency.

This leads to unstable and unpredictable outputs, making it difficult to achieve precise control over generation results.

From my experience, the issue seems related to either prompt adherence failure or context retention loss during processing.

I would appreciate confirmation on whether this is a known limitation, or if there are any recommended best practices or fixes to improve prompt consistency.

Hello @batler_platio,

This community dedicatedly supports the Gemini API and Google AI Studio.If you have issues with Google Flow, please visit the Google Flow App, click on the three dots in the menu, and select Feedback to report your issue.