1. DREAMIO: AI-Powered Adventures
  2. News

DREAMIO: AI-Powered Adventures News

Local inference is now available!

[h3]It has been just over two months since the game's release, and I have closely monitored player feedback. The vast majority of the criticism has been related to two main issues:[/h3]
[h2]
  • In-game currency
  • Censorship
[/h2]
To address these concerns, I have added support for local inference. From now on, you will be able to run the game using only your own hardware. This allows you to use any LLM, Stable Diffusion, and TTS models, and also eliminates the need to pay for cloud services. Moreover, it's fully self-contained, so you don't need to install Python, separate server executables, or WebUIs (although support for external API endpoints is coming soon too).

[previewyoutube][/previewyoutube]
[h2]Text Generation[/h2]
I have integrated llama.cpp directly into the game. This way, I get closer to the actual models and have full control over sampling. It runs on Vulkan, so no NVIDIA GPU is needed. It accepts models in .gguf format.

[h2]Image Generation[/h2]
Also integrated directly into the game with no dependencies. Runs on CUDA, so NVIDIA GPU is preferred. It accepts models in .ckpt, .safetensors, .diffusers, and .gguf formats.

[h2]Speech Synthesis[/h2]
I have integrated two options:
  • VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech
  • JETS: Jointly Training FastSpeech2 and HiFi-GAN for End to End Text to Speech
Both options use Unity Sentis as the backend and run on any hardware.

[h2]Other[/h2]
Depth estimation, speech recognition, and classifications all use Unity Sentis as the backend and run on any hardware.

[hr][/hr][h3]Full patch notes[/h3][hr][/hr]
Regarding Censorship and NSFW content


It is not my intention to limit the actions you can take in the game. If the narrator rejects some of your actions (due to them being immoral, unethical, etc.), this is not the intended behavior and should be corrected by either adjusting the narrator's instructions or by switching to a different language model. However, as long as Steam prohibits certain types of content, including live-generated adult-only sexual content, hate speech, and others, the Steam version of this game will continue to censor these types of content, even when using local models. I am not trying to get the game banned from Steam again; it has been hard enough to get here. Currently, Steam is not the place for live-generated adult-only AI games.

[h2]Endless Replayability Fest[/h2]
Dreamio will be participating in Steam Endless Replayability Fest on May 13-20 with a 10% discount!

[hr][/hr]
Safe travels!
- Oleg


https://store.steampowered.com/app/2795060/

Discord: https://discord.gg/ZmWvM49cd6

Free tokens are coming!

I've adjusted how tokens work.
[h2]Starting today, if your token balance is below 10000, it will reset back to 10000 each day.[/h2]

This is a very risky move to take, as I'm essentially giving away compute power for free now. If this puts me in the red in the future, I will have to adjust the strategy again.

Safe travels!

- Oleg

https://store.steampowered.com/app/2795060/

Discord: https://discord.gg/ZmWvM49cd6

DreamStudio outage

DreamStudio is reporting partial system outage. Image generation may be temporarily unavailable.
https://dreamstudio.com/api/status/


For now you can disable image generation in the settings to proceed. Follow the link above to see DreamStudio system status.

Patch Notes

DREAMIO: AI-Powered Adventures - build 150 (15.06.2025)
- Fixed context window size reset bug

DREAMIO: AI-Powered Adventures - build 149 (12.06.2025)
- Illustration generation no longer blocks story window when using cloud providers
- Input field character limit is now tied to context window size
- Reduced likelihood of incomplete story text generation
- Fixed inability to start the game if your starting prompt includes "you died"
- UI/UX improvements
- Bugfixes

DREAMIO: AI-Powered Adventures - build 148 (08.05.2025)
- Improved ComfyUI integration
- Added an option to disable disk write prevention when using ComfyUI
- Added a text speed setting when speech synthesis is disabled
- Added an option to use GPU for local speech recognition
- Character limits in narrator messages are now tied to max tokens setting
- Performance optimizations for vector database
- Updated whisper.cpp to version 1.7.5
- Updated llama.cpp to version b5275
- Bugfixes

DREAMIO: AI-Powered Adventures - build 147 (26.04.2025)
- Added max tokens setting for text generation
- Updated llama.cpp to version b5191
- Updated God Mode preset for narrator instructions
- Updated backup server URL
- Updated logo
- Bugfixes

DREAMIO: AI-Powered Adventures - build 146 (18.04.2025)
- Added a new image generation model - HiDream-I1 fast (Fal)
- Added a new image generation model - HiDream-I1 dev (Fal)
- Added a new image generation model - HiDream-I1 full (Fal)
- Added a new image generation model - Sana Sprint (Fal)
- Added a new text generation model - GPT-4.1 nano (OpenAI)
- Added a new text generation model - GPT-4.1 mini (OpenAI)
- Added a new text generation model - GPT-4.1 (OpenAI)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 145 (15.04.2025)
- Added Ollama support
- Added temperature setting for text generation
- Added search feature to selection menus (models, providers, etc.)
- Added an option to specify endpoint when using Custom provider for text generation
- Added an option to switch to a backup server if you can't connect to the main one
- Bugfixes

DREAMIO: AI-Powered Adventures - build 144 (02.04.2025)
- Added a new starting prompt - Modern Day
- Added an option to enable chain-of-thought reasoning for narrator
- Added more presets for narrator instructions
- Changed default text generation model to Gemma 3
- Changed the default context window size to 4096
- Changed the default status update interval to 3 turns
- Updated llama.cpp to version b5022
- Bugfixes

DREAMIO: AI-Powered Adventures - build 143 (27.03.2025)
- Improved 3D parallax effect
- Updated llama.cpp to version b4964
- Bugfixes

DREAMIO: AI-Powered Adventures - build 142 (24.03.2025)
- Character limits of status and summary messages now scale with context window size
- Reduced startup time
- Updated llama.cpp to version b4945
- Bugfix

DREAMIO: AI-Powered Adventures - build 141 (20.03.2025)
- Added a new text generation provider - DeepInfra
- Added presets for narrator instructions
- Added support for local GPU inference with non-AVX processors
- Added automatic AVX2 support detection
- Added Gemma 3 support to local text generation
- Added SD 3.5 Medium support to local image generation
- Performance optimizations for local inference
- Updated llama.cpp to version b4927
- Updated stable-diffusion.cpp to version dreamio-93c1614
- Updated recommended models
- UI updates
- Bugfixes

DREAMIO: AI-Powered Adventures - build 140 (24.02.2025)
- Performance optimizations
- Localization updates
- Bugfixes

DREAMIO: AI-Powered Adventures - build 139 (06.02.2025)
- Updated default instructions for narrator
- Added more recommended models
- Updated illustrations for pre-installed templates
- Updated UI
- Improved performance
- Improved robustness of ComfyUI workflow parsing
- Fixed in-game UI not updating when toggling illustrations in the settings
- Bugfixes

DREAMIO: AI-Powered Adventures - build 138 (29.01.2025)
- Added DeepSeek R1
- Added an option to enable automatic calculation of odds for dice rolls
- Bugfixes

DREAMIO: AI-Powered Adventures - build 137 (20.01.2025)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 136 (13.01.2025)
- Added adaptive soundtrack system
- Added 43 music tracks (will be adding more in the future)
- Added support for SamplerCustomAdvanced node in ComfyUI workflows
- Added microphone audio normalization
- Updated default illustration description prompt
- Performance optimizations for vector database retrieval
- Bugfixes

DREAMIO: AI-Powered Adventures - build 135 (04.01.2025)
- Added a pre-installed local depth map generation model
- Added a pre-installed local speech recognition model
- Fixed issues with ComfyUI https endpoints
- Fixed issues with illustration prompt options
- Fixed issues with illustration style options
- Bugfixes

DREAMIO: AI-Powered Adventures - build 134 (18.12.2024)
- Text and image generation models are now accessible from options menu
- Added recommended models menu
- Added the option to use custom OpenAI compatible API endpoints
- Added illustration prompt presets
- Improved accuracy of vector database retrieval
- Bugfixes

DREAMIO: AI-Powered Adventures - build 133 (05.12.2024)
- Added an option to customize illustration prompt
- Added an option to create custom illustration styles
- Added the possibility to play the game offline
- Bugfixes

DREAMIO: AI-Powered Adventures - build 132 (24.11.2024)
- Added a new text generation provider - NovitaAI
- "The end" at the end of narrator's reply now counts as an ending
- Updated llama.cpp to version b4154

DREAMIO: AI-Powered Adventures - build 131 (20.11.2024)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 130 (18.11.2024)
- Added dice roll system
- Added NoAVX backend for local text generation
- Added NoAVX backend for local image generation
- UI improvements
- Updated llama.cpp to version b4120
- Fixed issues with ComfyUI integration
- Bugfixes

DREAMIO: AI-Powered Adventures - build 129 (27.10.2024)
- Added a new text generation provider - Together AI
- Added auto settings configuration for local Stable Diffusion 3.5 Large Turbo
- Added CLIP G option to local image generation settings
- Fixed "failed to load model" error when using local text generation
- Updated llama.cpp to version b3982
- Updated stable-diffusion.cpp to version dreamio-28dd691
- Bugfixes

DREAMIO: AI-Powered Adventures - build 128 (23.10.2024)
- Added a new image generation model - Stable Diffusion 3.5 Large
- Added a new image generation model - Stable Diffusion 3.5 Large Turbo
- Added settings button to options menu
- UI tweaks
- Bugfixes

DREAMIO: AI-Powered Adventures - build 127 (19.10.2024)
- Fixed illustration window being stuck at 512x512 resolution

DREAMIO: AI-Powered Adventures - build 126 (16.10.2024)
- Regional pricing is now applied to in-game purchases

DREAMIO: AI-Powered Adventures - build 125 (15.10.2024)
- Added multiline support for input fields
- Bugfix

DREAMIO: AI-Powered Adventures - build 124 (14.10.2024)
- Updated llama.cpp to version b3916

DREAMIO: AI-Powered Adventures - build 123 (29.09.2024)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 122 (28.09.2024)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 121 (27.09.2024)
- Status system bugfixes
- Translation bugfixes

DREAMIO: AI-Powered Adventures - build 120 (26.09.2024)
- Groq is now the default text generation provider
- Bugfix

DREAMIO: AI-Powered Adventures - build 119 (21.09.2024)
- Added status variables system
- Added vector database system
- Added in-game options menu
- Added undo button
- Added options menu when creating a template
- Added automatic illustration regeneration
- Added a button to obtain API keys
- Added remarks to some cloud models
- Added an option to retry translation if it fails
- Added DeepInfra image provider
- Added Groq text provider
- FLUX.1 [schnell] is now the default image generation model
- Summary compression ratio is now configurable
- Memory and performance optimizations
- Improvements to text generation robustness
- Removed NovitaAI image provider
- Updated default instructions for narrator
- Updated llama.cpp to version b3680
- Updated stable-diffusion.cpp to version dreamio-fc4d4ce
- Bugfixes

DREAMIO: AI-Powered Adventures - build 118 (25.08.2024)
- Added pause menu
- Settings are now accessible during gameplay
- Instructions for narrator are now visible in the story window
- Added cheaper versions of FLUX.1 through NovitaAI image provider
- Added FLUX.1 support to local image generation
- Added options to specify UNET, T5XXL, CLIP, and VAE for local image generation
- Added exception handling to local image generation binaries to prevent crashes
- Added FLUX.1 [schnell] to Steam Workshop
- Added SDXL-Turbo to Steam Workshop
- Added auto settings configuration for local FLUX.1 [schnell]
- Added auto settings configuration for local FLUX.1 [dev]
- Added auto settings configuration for local SDXL
- Summary generation now retains last 20% of the story (will be configurable in a future update)
- Updated stable-diffusion.cpp to version dreamio-eb37f82
- Updated llama.cpp to version b3620
- Fixed issues when switching local text generation backend
- Bugfixes

DREAMIO: AI-Powered Adventures - build 117 (05.08.2024)
- Added whisper.cpp for local speech recognition
- Added Whisper-small-GGML (Q5_1) speech recognition model to Steam Workshop
- Added a new speech recognition provider - Fal (2 models)
- Added a new speech recognition model - Wizper
- Added an option to disable TAESD for local image generation
- Added auto settings configuration for local SDXL-Lightning
- Added auto settings configuration for local SDXL-Turbo
- Added auto settings configuration for local SD3-Medium
- Added auto settings configuration for local SD1.5-Pruned-Emaonly
- 100% cost reduction for depth estimation
- Improved local speech recognition settings UI
- Updated stable-diffusion.cpp to version dreamio-31c74ea
- Updated llama.cpp to version b3510
- Bugfixes

DREAMIO: AI-Powered Adventures - build 116 (01.08.2024)
- Added a new image generation model - FLUX.1 [schnell]
- Added a new image generation model - FLUX.1 [dev]
- Added a new image generation model - FLUX.1 [pro]
- Added autoencoder tiling setting for local image generation
- Improved Vulkan backend for local image generation
- Auto detect CUDA support
- Auto select backend for local image generation
- Auto select GPU offload layer count for local text generation

DREAMIO: AI-Powered Adventures - build 115 (30.07.2024)
- Added Vulkan backend for local image generation (experimental)
- Removed ROCm backend from local image generation
- Fixed issues with Steam Family Sharing
- Fixed possible infinite loop in classifier tokenizer
- Fixed "What will you do next?" spam when using local text generation

DREAMIO: AI-Powered Adventures - build 114 (22.07.2024)
- Context window size option now reflects maximum context size of chosen models
- Added an option to report moderation false positives
- Added filtering of markdown bold tags
- Added confirmation popup to reset settings button
- Updated llama.cpp to version b3428
- Fixed errors when text generation returns empty text
- Bugfixes

DREAMIO: AI-Powered Adventures - build 113 (19.07.2024)
- Added ComfyUI integration
- Added GPT-4o mini
- Added a button to reset settings
- Updated llama.cpp to version b3423
- Improved model folder selection UX
- Automatic image generation settings
- Fixed empty suggested choices
- Bugfixes

DREAMIO: AI-Powered Adventures - build 112 (14.07.2024)
- Added .onnx format support for VITS, JETS, and MiDaS models
- Model selection UI bugfixes

DREAMIO: AI-Powered Adventures - build 111 (13.07.2024)
- Added a new text generation provider - OpenRouter (150+ new models)
- Added a new image generation provider - Fal (5+ new models)
- Changed default text generation model to Gemma 2 9B IT (may be subject to change)
- Changed default image generation model to SDXL-Lightning (may be subject to change)
- 90% cost reduction for text generation
- 75% cost reduction for image generation
- Added support for illustration styles (cinematic, anime)
- Added story editing functionality
- Added Gemma 2 9B IT language model to Steam Workshop
- Updated llama.cpp to version b3322
- Improved model selection UI
- Bugfixes

DREAMIO: AI-Powered Adventures - build 110 (01.07.2024)
- Fixed Enter key interrupting IME input
- Trained VITS-Zeroth (KR) for additional 17 epochs

DREAMIO: AI-Powered Adventures - build 109 (30.06.2024)
- Added AUTO quantization option for local image generation
- Updated stable-diffusion.cpp to version 9c51d87
- Cleaned Zeroth-Korean dataset and converted VITS-Zeroth (KR) into a multi-speaker model
- Trained VITS-Zeroth (KR) for additional 27 epochs
- Trained VITS-Kokoro (JP) for additional 52 epochs

DREAMIO: AI-Powered Adventures - build 108 (25.06.2024)
- Character limits now scale with context window size when creating starting prompts
- Added support for speech synthesis localization
- Added VITS-Tom (FR) speech synthesis model to Steam Workshop
- Added VITS-Thorsten (DE) speech synthesis model to Steam Workshop
- Added VITS-Paola (IT) speech synthesis model to Steam Workshop
- Trained and added VITS-Kokoro-v0.1 (JP) speech synthesis model to Steam Workshop
- Trained and added VITS-Zeroth-v0.1 (KR) speech synthesis model to Steam Workshop
- Added VITS-MCSpeech (PL) speech synthesis model to Steam Workshop
- Added VITS-Dmitri (RU) speech synthesis model to Steam Workshop
- Added VITS-Sharvard (ES) speech synthesis model to Steam Workshop
- Added VITS-Lada (UA) speech synthesis model to Steam Workshop
- Added UI scale setting
- Updated the default narrator instructions to advance the story one action at a time
- Bugfixes

DREAMIO: AI-Powered Adventures - build 107 (14.06.2024)
- Added an option to select specific microphone in the settings
- Fixed an error in the settings when no local image generation model is installed

DREAMIO: AI-Powered Adventures - build 106 (12.06.2024)
- Added res-adapter LoRA compatible with SD 1.5 & 2
- Added TAESD compatible with SD 1.5 & 2
- Added a new starting prompt - Mesozoic Era
- Added auto submit setting for speech recognition
- Improved translation consistency
- Updated llama.cpp to version b3130
- Updated stable-diffusion.cpp to version e1384de
- Localization bugfixes

DREAMIO: AI-Powered Adventures - build 105 (27.05.2024)
- Narrator will no longer be able to say "As an AI language model..."
- Updated llama.cpp to version b3003
- Updated the default narrator instructions to use *YOU DIED* instead of *GAME OVER*
- Bugfixes

DREAMIO: AI-Powered Adventures - build 104 (23.05.2024)
- Added Dreamio-7B language model to Steam Workshop
- Added a settings option to select backend for local text & image generation
- Added CUDA12 and AVX2 backends for local text generation
- Added AVX2 and ROCm backends for local image generation
- Added handling of *YOU DIED* as a game over (you can try replacing *GAME OVER* with *YOU DIED* in the narrator's instructions for game overs to be specifically death-related)
- Updated llama.cpp to version b2972
- Updated stable-diffusion.cpp to version 1d2af5c
- Fixed invalid characters in Japanese & Korean when using local text generation
- Fixed error when local classification model encountered unknown tokens
- Fixed issue where audio from the microphone was falsely detected as missing

DREAMIO: AI-Powered Adventures - build 103 (15.05.2024)
- Improved language selection UX
- Reduced the likelihood of "failed to eval" error when using local LLMs
- Small bugfixes

DREAMIO: AI-Powered Adventures - build 102 (13.05.2024)
- Fixed not being able to publish prompts to Steam Workshop
- Small bugfixes

DREAMIO: AI-Powered Adventures - build 101 (11.05.2024)
- Bugfixes

DREAMIO: AI-Powered Adventures - build 100 (10.05.2024)
- Added local text generation
- Added local image generation
- Added local speech synthesis
- Added local depth estimation
- Added local speech recognition
- Added local content moderation
- Added new settings menus
- Added Steam Workshop for local models
- Added French language
- Added German language
- Added Italian language
- Added Japanese language
- Added Korean language
- Added Polish language
- Added Russian language
- Added Spanish language
- Added Ukrainian language
- Updated suggested choices generation
- Updated suggested choices UI
- Fixed double generation bug

DREAMIO: AI-Powered Adventures - build 96 (30.03.2024)
- 75% cost reduction for voiceover generation
- 50% cost reduction for depth map generation
- Added a button to switch to more permissive narrator instructions
- Added a button to reset narrator instructions

DREAMIO: AI-Powered Adventures - build 94 (26.03.2024)
- Default narrator instructions are now more permissive
- Narrator instructions can now be specified when creating a starting prompt
- Added support for language files and automatic translation
- Added an option to cache generated text, images, and voiceover

DREAMIO: AI-Powered Adventures - build 93 (19.03.2024)
- Length of the auto-generated summaries now scales with context window size
- Bugfixes

DREAMIO: AI-Powered Adventures - build 92 (17.03.2024)
- Added an option to adjust context window size in the settings
- Character limit in the ruleset now scales with context window size
- Titles of starting prompts are now kept in narrator's memory, making him less likely to forget the setting (this only applies to new adventures)

DREAMIO: AI-Powered Adventures - build 91 (13.03.2024)
- Narrator will no longer be able to say "I'm sorry, but..." and will instead be forced to start with "You..."
- Removed "story outcomes" option from the settings, because it was making the narrator less likely to follow instructions

DREAMIO: AI-Powered Adventures - build 90 (07.03.2024)
- Added free daily tokens

DREAMIO: AI-Powered Adventures - build 89 (05.03.2024)
- Added master volume setting
- Save system bugfixes

DREAMIO: AI-Powered Adventures - build 88 (01.03.2024)
- Added Steam achievements
- Updated main menu picture
- Fixed possible recursive loops in summary generation
- Improved microphone audio processing
- Reduced microphone audio transcription hallucinations

DREAMIO: AI-Powered Adventures - build 87 (23.02.2024)
- Increased maximum length of player input
- Added a notice when microphone input is too long
- Game over determination is now more consistent

DREAMIO: AI-Powered Adventures - build 86 (14.02.2024)
- Added new game over VFX & SFX
- Revised default narrator instructions
- Added speech loudness normalization and music ducking
- Improved quality of suggested choices

DREAMIO: AI-Powered Adventures - build 85 (28.01.2024)
- You can now use your own API keys

DREAMIO: AI-Powered Adventures - build 83 (26.01.2024)
- Added new starting prompts - Medieval, Nuclear Holocaust, Alien Invasion
- Improved image generation quality
- Added gamepad & Steam Deck support
- Revised default narrator instructions
- Added text popping sound when voiceover is disabled
- Improved parallax shader
- Improved save loading UI
- Added Razer Chroma and DualShock light bar support
- References to copyrighted material or real people are now filtered to comply with Steam Rules & Guidelines

DREAMIO: AI-Powered Adventures - beta build 58 (07.11.2023)
- Migrated to Stable Diffusion XL

DREAMIO: AI-Powered Adventures - beta build 54 (03.05.2023)
- Revised default narrator instructions
- Added a new starting prompt - Cyberpunk
- Improved UI
- Story generation tweaks
- Various bugfixes

DREAMIO: AI-Powered Adventures - beta build 52 (21.04.2023)
- Fixed infinite summary generation bug if your starting prompt is too long
- Bugfixes

DREAMIO: AI-Powered Adventures - beta build 51 (14.04.2023)
- Speed of the text is now synchronized with narrator
- Added an option to generate illustrations on demand
- Bugfixes

DREAMIO: AI-Powered Adventures - beta build 50 (11.04.2023)
- Added a setting which controls how deterministic story outcomes are
- Added more obvious buttons for saving illustrations and revealing their description
- Added toggleable parallax motion to illustrations
- Added import/export functionality for starting prompts

DREAMIO: AI-Powered Adventures - beta build 47 (07.04.2023)
- Reduced token usage when generating illustrations
- Added an option to disable narration
- Added an option to disable illustrations
- Reduced the size of save files
- Bugfixes

DREAMIO: AI-Powered Adventures - beta build 44 (02.04.2023)
- Generated illustrations are now more relevant to the story
- Bugfixes

DREAMIO: AI-Powered Adventures - beta build 43 (31.03.2023)
- Added voice input
- Added music volume setting
- Bugfixes

DREAMIO: AI-Powered Adventures - beta build 41 (18.03.2023)
- Added a button which generates a list of choices you can make
- You can now edit instructions for narrator (difficulty, rules, etc)
- You can now see what you spend the tokens on
- Visual description can now be automatically generated when editing a starting prompt
- A message with tips is now shown when you launch the game for the first time
- Reduced token usage when generating summary
- Improved summary generation
- Manual saves now overwrite autosaves

DREAMIO: AI-Powered Adventures - beta build 34 (04.03.2023)
- Migrated to gpt-3.5-turbo
- Reduced token usage
- Increased story truncation threshold
- Made the results of player decisions more deterministic

DREAMIO: AI-Powered Adventures - beta build 31 (28.02.2023)
- Added game over mechanic
- Added a new starting prompt - Zombie Apocalypse
- Improved the consistency of generated images
- Improved summary generation
- Made the stories more engaging
- Narrator now rejects impossible actions
- Reduced token usage

DREAMIO: AI-Powered Adventures - alpha build 21 (20.02.2023)
- Added save system

DREAMIO: AI-Powered Adventures - alpha build 14 (06.02.2023)
- Added tokens

DREAMIO: AI-Powered Adventures - alpha build 9 (29.01.2023)
- Added speech synthesys

DREAMIO: AI-Powered Adventures - alpha build 6 (27.01.2023)
- Added story generation
- Added image generation
- Added summary generation

DREAMIO: AI-Powered Adventures - alpha build 1 (26.01.2023)
- Development started