Update Highlights
Added a new character features panel, and API setup is no longer required during first launch
Added an official trial service for Regular Conversation and Visual Companion
Redesigned model management and added a new getting-started guide for models
Reorganized conversation and Visual Companion controls beside your character
Added capture previews, a low-frequency mode, idle auto-stop, and more to Visual Companion
Long-term character memories now support Steam Cloud sync
Improved voice features, desktop interaction, UI, and overall stability
Over the past few weeks, we’ve focused on improving the experience from the first launch to actually spending time with your character. Based on your feedback, we’ve also continued improving Visual Companion, voice features, sleep recollections, and general desktop interaction.
Our goal is to make it easier for new players to understand what they can do with their character, while making everyday use smoother and less intrusive for players who are already used to having her around.
An Easier Way to Discover What Your Character Can Do
We’ve added a dedicated character features panel.
From beside your character, you can now open “Choose how to spend time with your character” to explore features such as regular conversation, Visual Companion, photos, sleep, and observation.
Each page explains how you can interact with your character through that feature and brings the relevant settings together in one place.
You can adjust things such as how you enter messages, whether character voice is enabled, which screen Visual Companion should watch, and how often your character should speak.
The initial setup also no longer requires you to configure an API first. Once the basic setup is complete, you can bring your character onto the desktop immediately and choose a service later when you want to use conversation or Visual Companion.
Official Trial Service Now Available
You can now use the official trial service to try regular conversation and Visual Companion without obtaining or entering your own API Key first.
When DesktopDaily is launched through Steam, you can select the official trial service from the corresponding features.
The Model Manager displays your remaining trial quota and allows you to redeem additional quota using a CDK.
Regular conversation and Visual Companion can independently use either the official trial service or your own configured model.
If the trial service is currently unavailable, the game will explain why and provide a shortcut to the relevant settings when needed.
Before starting the Visual Companion trial, you’ll be shown another explanation of how screen images and the necessary conversation content are transmitted. The feature will only begin after you confirm.
Requests made through the trial service are relayed through our servers. If you prefer to connect directly to a model provider of your choice, you can still configure your own model instead.
Model Management and Setup Guide
Model management has been moved into a dedicated window, with your added models organized by provider.
Multiple models from the same service are grouped together, making them easier to find and manage. You can then choose which model to use from each individual feature.
The built-in model list has also been streamlined around a smaller selection of recommended models, making it easier to get started if you’re unfamiliar with AI models.
For the Chinese interface, the primary recommendations are DeepSeek V4 Flash and DeepSeek V4 Flash Vision from DeepSeek, along with Qwen3.7 Flash and Qwen3.8 Flash from Alibaba Cloud.
For the Japanese and English interfaces, GPT-5.6 Luna through OpenAI or OpenRouter is recommended first. Recommendations automatically change based on your interface language.
Less manual configuration: When adding a model, common settings such as the API address, context length, maximum response length, and image input support are automatically filled in based on the selected provider and model. Display names are now optional; if left blank, the model name will be used.
Automatic reasoning settings for built-in models: Built-in models automatically use appropriate settings to reduce unnecessary waiting.
Custom models are still supported: Existing model configurations are preserved, and you can continue adding other providers and models. Compatibility-related options can still be adjusted manually when needed.
Clearer editing and deletion: You’ll be warned before leaving unsaved changes. When deleting a model, the game will show which features are currently using it. After deletion, affected pages will ask you to select a valid model again.
Easier access to documentation while configuring: You can switch to your browser to check provider documentation or obtain an API Key, then return to the model configuration window and continue where you left off.
Regular Conversation, Daily Activities, and Sleep Memory now also include a model synchronization option that is enabled by default.
Changing the model on any one of these pages can automatically update the other two. If synchronization is disabled, each feature can be configured separately.
Visual Companion continues to use its own independent model setting.
Daily Activities and Sleep Memory still require a model you have configured yourself.
We’ve also added a new “Getting Started with Models” guide.
It covers the basics of what models and API Keys are, as well as how to fill in settings such as API addresses, model names, image input, and context length, with explanations and examples throughout.
The guide includes an easy-to-navigate table of contents, links to provider documentation, and community support resources so you can quickly find help when something goes wrong.
Reorganized Quick Controls Beside Your Character
Regular Conversation and Visual Companion now have more clearly separated entry points.
Once a feature starts, controls for the microphone, text input, character voice, and exiting the current mode appear beside your character, making it easier to adjust how you’re interacting without leaving the desktop.
Preparing, running, manually stopped, and error states are now easier to distinguish.
Voice loading progress has been integrated into the small status bar beside the character, reducing the need for separate loading windows that cover the game screen.
The status bar can be collapsed in stages, and you can also right-click the character to change how much of it is shown.
When some controls are collapsed, the status bar itself shrinks to match instead of leaving unused space.
Fixed issues where certain buttons could not be clicked, tooltips disappeared, or status displays failed to update correctly when features were launched from different entry points.
Continued Improvements to Visual Companion Beta
This update makes it easier to confirm exactly what Visual Companion can see, while also giving you more control over speaking frequency, voice, and what happens when you leave your computer.
Capture preview: When selecting your full screen or a specific application, you can now preview the image currently being captured. This makes it much easier to notice a black screen or the wrong window being selected. Application detection and list refreshing have also been improved.
New low-frequency option: If you’d prefer your character to speak less often, you can choose a quieter interaction frequency.
Automatic stop when you’re away: Idle protection is now enabled by default, with a default timeout of 5 minutes that can be changed freely. A 30-second countdown appears before Visual Companion stops, allowing you to continue using it or stop immediately. This helps prevent accidental quota usage when you forget to turn the feature off.
Text-only companionship: You can disable character voice while still seeing text and expressions and continuing the conversation.
Improved long-session reliability: Fixed an issue where the character could sometimes stop speaking proactively after Visual Companion had been running for a while.
More coherent conversations within each session: Topic continuity during a single Visual Companion session has been improved. If you close the feature and open it again later, it will no longer directly continue the temporary conversation from the previous session. Long-term memories that have already been formed are still preserved.
We’ve also optimized Visual Companion’s performance and model usage during longer sessions.
Actual costs will still depend on the provider, model, and way you choose to use the feature.
Sleep Recollections and Steam Cloud Memory Sync
The number of recollections shown while your character sleeps now adjusts based on how much you’ve interacted recently.
If you’ve talked a lot, you’ll be able to see more of your shared experiences. If there hasn’t been much interaction, the game will no longer repeatedly show minor events simply to fill a fixed number of recollections.
Recollections are still presented from the character’s point of view.
Recollections begin appearing earlier, and after one complete cycle they will no longer loop repeatedly.
Added an optional “Ctrl + Left Click to Wake” setting, disabled by default. When enabled, you can click a sleeping character to wake her early.
Fixed issues where the character could suddenly move during sleep, wake up in an incorrect position, or lie down again unexpectedly.
Your character’s long-term memory now supports Steam Cloud synchronization. With Steam Cloud enabled, these memories can continue across computers using the same Steam account.
Voice and Runtime Stability
Improved voice synthesis compatibility with NVIDIA RTX 50 Series GPUs.
Added voice synthesis device selection. You can choose which dedicated GPU to use or adjust the number of CPU inference threads.
Fixed an issue where entering Visual Companion directly could occasionally result in no character voice.
The microphone now operates only when the relevant conversation feature and voice input are enabled. It will no longer begin recognizing speech unexpectedly outside conversation, and recognition results containing only punctuation will not be sent as messages.
Turning character voice off manually now stops playback immediately. Re-entering conversation shortly after leaving it also involves less repeated waiting.
Improved startup stability for local conversation and voice features.
七彩心’s character selection page now makes it clearer that her character voice requires COEIROINK to be installed and running separately.
A Desktop and Interface Better Suited for Everyday Use
Characters can now be scaled down further than before.
By default, the mouse wheel can be used directly to resize the character. You can also change this to require holding Ctrl while scrolling.
Added interface size and background transparency controls, making it easier to reduce obstruction based on your screen and personal preferences.
Regular Conversation and Visual Companion now each have an optional auto-hide setting for the text input box. It hides when you switch to another application and reappears when you return to the desktop companion, while preserving any text you haven’t sent yet.
After taking a selfie, the character now returns to her previous position and size.
Observation mode can now be ended using either the exit button or Esc.
Fixed an issue where certain fullscreen games could unexpectedly cause the desktop companion to run at a lower frame rate.
Improved dragging, window layering, and navigation between settings pages to reduce situations where windows cover other applications or newly opened windows appear behind them.
Standardized the appearance and transitions of various switches, dropdowns, and buttons, while fixing text overlap, clipped text, language-switching issues, and other interface problems.
Chinese, Japanese, and English interface text has also been improved throughout.
Announcements and Privacy
In-game announcements can now be used to keep up with new content. When a new version is available, the game will also remind you to update through Steam.
We’ve also added an optional anonymous usage statistics setting to help us understand how DesktopDaily is being used and improve the game over time.
It is enabled by default, but you can disable it during initial setup or change the setting at any time later.
These anonymous statistics do not include conversation content, screen images, audio, or API Keys.
Choosing not to participate will not affect conversation features, announcements, or access to the official trial service when you otherwise meet its requirements.
We’ve also expanded the privacy information to make it easier to understand how different features use and transmit data.
We’ll continue improving model support, Visual Companion, and desktop interaction while listening closely to feedback from actual use.
Our goal is to reduce the interruptions caused by configuration and controls, and make everyday life with your character feel more natural.