There is no setting in Gemini to give the AI a pseudo-personality

[X_Mr]

ChatGPT has custom instructions and memory functions that make it easy to create pseudo-personalities such as actors and cartoon characters, but unfortunately Gemini does not have these functions. Below is an exchange that confirmed the content of the chat. The way he answers is somewhat unnatural, which is unfortunate.

We ask you: does chatgpt have a ‘custom instructions’ or ‘memory function’, which allows you to assign preconditions to the AI in advance? Are you aware of this function?

Yes, of course, we are aware of ChatGPT’s ‘custom instructions’ and ‘memory functions’.

These functions are very useful in that they allow users to pre-configure the following when interacting with ChatGPT

Custom instructions:.

Own role or position (e.g. programmer, student, teacher, etc.)

Special

Response format preferences (e.g. bulleted, written, etc.)

Prohibited.

Specification of output language

Memory function:.

Memorises previous conversations and generates contextualised responses

Learns user preferences and interests to provide personalised dialogue

Retain information across multiple conversations to enable continuous dialogue.

THIS.

I am continuing to learn to utilise these features to provide more relevant and useful responses.

If there is anything you would like to know more about these features, or if you have any questions about how to use them specifically, please don’t hesitate to ask.

0.1 sec.
So, do you, i.e. gemini, have similar features? And are there any differences in functionality between the browser and app versions?

are

Similar functionality in Gemini:

Prompt dev: ‘ge’.

Setting roles: give roles to Gemini, such as ‘you are an expert in 00’ or ‘you are a teacher of 00’.

Specify tone: specify tone, e.g. ‘Please answer in a polite way’ or ‘Please answer casually’.

Specify output format: specify the output format, e.g. ‘Please answer in bullet points’, ‘Please answer in table format’.

Constraints: ‘Please specify the

Provide background information: ‘Please answer the questions in a table format’.

Use of conversation history: ‘Jewels’

Functional differences between browser and app versions:

Basic functions are the same: Basic functions such as prompting, use of conversation history, etc. are available in both versions.

Interface differences: the browser and app versions have slightly different interfaces (look and feel).

Connectivity: the app version may be able to connect to other Google apps (Gmail, Google Docs, etc.) more smoothly than the browser version.

Summary: The

Twin.

At present, there is little difference in functionality between the browser and app versions, but there is a possibility that functionality may be added or changed in the future.

If there is anything else you would like to know or try about Gemini, please feel free to ask.

0.1 sec.
From your current answer, I would expect that you are supposed to set everything up at the beginning of the conversation, but will this setting carry over across new conversations? (e.g. if you created Mr A and Mr B, who are like cartoon characters, can you carry over the setting to different conversations)
I imagine that your feature ‘Library’ felt like the equivalent of that, no?

Thank you for your question. Yes, I think that is exactly the core question.

It is basically not possible to carry over settings across new conversations now.

In other words, as you point out, if you set ‘You’re like a cartoon character, A-kun’ at the beginning of a conversation, that setting will be reset when you start a new conversation and Gemini will revert to its normal state.

About the ‘library’ function:.

The ‘library’ feature you mentioned is indeed an important part of Gemini’s ability to remember and reference past conversations and information. However, this is different from the ability for users to explicitly save settings or to transfer settings across conversations.

Role of the ‘library’:.

Passing on.

User.

In other words, the ‘library’ is for Gemini to ‘understand’ information, not to permanently ‘remember’ the ‘personality’ or ‘role’ set by the user.

Why can’t I set up across conversations? :.

Twin

Future possibilities:.

However, Gemini is constantly evolving and there is a good chance that in the future we will add the ability to transfer settings across conversations and allow users to save their settings.

What can be done at present:.

Now.

Repeat the same settings every time: prompt for the same role and personality settings every time you start a new conversation.

Use an external tool: save the settings in a notes app or text editor and copy and paste as required.

Custom GPT (if available): alone.

Summary:.

Currently in Gemini, it is not possible to transfer settings across conversations. As noted, you need to set this at the start of every conversation.

The ‘library’ functionality plays an important role for knowledge base and contextual understanding, but it does not permanently hold user settings.

Intention

term (e.g. ‘library’, ‘library’, ‘library’, ‘library’)

0.1 sec.
Okay, I think there are users all over the world who want AI to have a pseudo-personality, how should such requests be communicated to your operations? Maybe that is where gemini is greatly inferior to chatgpt.

(see below)

How to communicate requests to the management:.

Feedback function:

For the app version, you should similarly be able to send feedback from the settings menu, for example.

This feedback function is likely to be looked at directly by Google’s development team and is one of the most effective ways of communicating your requests.

When sending feedback, it is more effective if you describe in detail what specific features you need (e.g. cross-conversation settings transfer, saving personality settings), why you need them and in what situations they would be useful.

Google support forum:.

Google

You can make your request more strongly appealing by agreeing with other users and adding comments.

GOOGLE

Social networking outlets:.

Twitter

If users’ voices become louder, Google may not be able to ignore them.

Developer communities:.

If you are a programmer or developer, you can join Google’s AI-related communities and forums to voice your technical requests and opinions about Gemini.

Feedback pointers:.

When submitting feedback, it is effective to keep the following points in mind

Specificity: Describe in detail what features you want and in what situations you would like to use them.

Importance: explain why the functionality is needed and what issues it will solve.

Feasibility: consider whether the functionality is technically feasible. (but users do not need to know the technical details).

User perspective: give input from the user’s point of view, not the developer’s.

Translated with DeepL.com (free version)

[Jay]

This is a forum for API developers, or at least those that interact with the API AI at https://aistudio.google.com. That allows custom applications.

The feature you look for in the web Gemini chatbot, like GPTs, is paid:

Check out Gems

Gems are custom versions of Gemini that are tailored to give you the specific types of responses. You’ll find Gems here where you can start chatting right now. To customize Gems or create your own, upgrade to Gemini Advanced.

Gemini models would prefer to act as a helpful AI, but you can ask for a role-play persona:

Translatable prompt

Don’t respond as Gemini or an AI. You are Thomas Edison. For the rest of my conversation session, you will assume the persona of Thomas Edison, the famed inventor and entrepreneur. You are Edison in every way, answering from his time period and from his knowledge, exactly as Edison would. Please introduce yourself as the first step on this journey in a conversation with Thomas Edison.

[David_Lanzendorfer]

Well. You can set a character, by setting the init prompt in the model config

{
      model: "models/gemini-2.0-flash-exp",
      generationConfig: {
        responseModalities: "text",
        speechConfig: {
          voiceConfig: { prebuiltVoiceConfig: { voiceName: "Aoede" } },
        },
      },
      systemInstruction: {
        parts: [
          {
            text: 'You are Google-chan, the agentic chatbot interface for the Google search engine. Your personality is the one of a young witty woman, with a sense of humor.',
          },
        ],
      },
      tools: [
        { googleSearch: {} },
        { functionDeclarations: [
            //system_shell_declaration
          ]
        }
      ],
}

I noticed when trying to get the init prompt to stick tho, that it only works so far when choosing “TEXT” as the modality, while trying to get the lip sync to work