FlowyAIPC
  • English

    • Virtual Character Chat

    Define an AI persona with custom personality traits, then connect it to WeChat to chat with it from your phone as if it were a friend.

    Create a Virtual Character

    1. Click Create Virtual Character to open the configuration wizard.

    Step 1 — Basic Information

    Fill in the basic profile, then click Next:

    • Name — what the character is called
    • Gender / Role — male, female, or animal
    • World Setting — modern, fantasy, ancient, sci-fi, etc.
    • Persona, Tone, and Boundaries — short description of the character's personality and limits

    Step 2 — Persona

    Define the character's personality, then click Next:

    • Personality
      • Temperament
      • Relationship to the user
    • Tone & Style
      • Rational vs. emotional
      • Proactive vs. reserved
      • Casual vs. formal
    • Knowledge & Memory
      • Structured system prompt
      • Few-shot examples
    • Values & Guidelines
      • Whether to offer unsolicited suggestions
      • Whether to use emoji
      • Whether to allow small talk

    Step 3 — Capabilities

    Select the capabilities the character should have, then click Next:

    • Text Chat — base capability; powered by the locally installed LLM
    • Image Generation — optional; powered by the local text-to-image model. Verify image generation works (see the Image Generation chapter) before enabling this.
    • Voice — optional; not yet supported

    Step 4 — Model Strategy

    1. Choose a Text Chat Model (required).
    2. If you enabled image generation in Step 3, also choose an Image Generation Model.
    3. Click Next.

    Step 5 — Link to WeChat

    1. Scan the QR code with WeChat to link your account.
    2. After linking, a Linked Successfully confirmation appears.

    Step 6 — Confirm

    1. Review the full configuration summary.
    2. Click Confirm to finalize and connect the character to WeChat.

    Chatting in WeChat

    Use it like any other WeChat contact.

    • Send messages to chat normally.
    • Request images when text-to-image is enabled.

    FAQ

    Q: Can I use this feature offline?

    • The character uses a local model in Herdsman, so the AI processing itself does not depend on the cloud. However, WeChat itself requires an internet connection on your computer for messages to be relayed.

    Q: Are there any risks with the WeChat QR-code link? Could it expose information or initiate transfers?

    • No. There are no such risks:
      • Local model processing keeps your data on-device.
      • The integration uses WeChat's official APIs and follows WeChat's security requirements.
      • The character cannot access your money or transfers — think of it as an AI friend you can chat with. It cannot remotely control your phone, access stored data, or initiate financial actions.

    Q: Why can't my character generate images?

    • Image generation depends on the local text-to-image model, which is hardware-intensive. Verify the Image Generation chapter requirements first.
    • If hardware is sufficient but it still fails, you likely have too many models running and not enough resources. Stop the models you aren't using: Models page → review running models → stop the ones you no longer need.

    Q: Why can't my character understand images I send?

    • Image understanding requires a multimodal model. Confirm that the text chat model you selected is multimodal: Edit configuration → Model Strategy → check the model type → switch to a multimodal model → continue through the wizard to save.