FriedrichAI: Offline AI news & patch notes

FriedrichAI just got a major usability pass.

This update adds a cleaner app-style interface with better spacing, a more focused chat area, readable local/model/backend status indicators, and a simplified tool bar so the main window feels less like a debug panel and more like an actual workspace.

The bigger addition is the new Kanban Project Board. You can now open a board view from the sidebar and organize tasks into:

Todo Doing Blocked Done

The board is meant for local project planning and task organization. You can add tasks, refresh the board, and move cards between columns without leaving the app.

Also included:

• Cleaner sidebar navigation

• Collapsed tool drawer to reduce button clutter

• Working saved-chat search

• Better default window sizing

• Debug warning cleanup

• Continued offline/local-first behavior

As always, FriedrichAI still runs locally. No cloud requirement, no required API key, and no tracking.

This is part of the ongoing Early Access polish pass. The goal is simple: make FriedrichAI easier to use every day, not just technically functional.

This update continues the push toward making FriedrichAI feel less like a single chat window and more like a local AI workspace.

The newest testing build adds two major framework pieces:

Model management lets FriedrichAI detect local GGUF models, switch selected models, reset back to the default bundled model, and restart cleanly after a model change. This is still local-first: users can add models inside the FriedrichAI workspace without needing cloud accounts or API keys.

Saved chat management adds separate local chats. You can create chats, switch between them, reopen earlier work, and summarize the active chat. The backend now saves chat state locally and reloads useful context when a chat is opened.

Current command examples:

There is also an early UI layer for these controls so users do not need to type every command manually.

Next up is the Kanban/work board system. The goal is simple: let FriedrichAI track tasks, project steps, blockers, and next actions locally, then connect that board to the existing project-reading and chat systems.

This is still Early Access framework work, but the direction is clear:

FriedrichAI is moving toward a private local workspace assistant, not just a chatbot.

Hey guys, a lot has been changing, being added. I am going to be adding as much detail as I can on how to use everything in the FAQ soon. thank you.

This update focuses on making FriedrichAI more stable, predictable, and ready for the next layer of local workflow tools.

The biggest change is language handling. FriedrichAI now defaults to English instead of trying to auto-detect the reply language from every message. Auto-detect sounded useful, but in real use it could cause unwanted language switching when users pasted mixed text, foreign words, or examples. Explicit language requests still work, so you can still ask FriedrichAI to reply in Spanish, French, German, Portuguese, Russian, or other supported languages when needed.

This patch also includes additional cleanup for repeated or echoed response text. Short answers should be less likely to repeat themselves or mirror the user’s prompt back awkwardly.

Optional web access has also been added as a user-controlled feature. FriedrichAI remains offline by default. Web access is off unless the user enables it and provides their own API key. This is meant for current-information lookups and research-style assistance while keeping the normal FriedrichAI experience local-first.

This update also includes early internal hooks for future board/task workflow features. The full Kanban-style workflow system is not live yet, but this patch lays some of the groundwork for it.

Included in this update:

- English is now the default response language

- Auto language switching disabled by default

- Explicit language requests still supported

- Improved repeated-response cleanup

- Optional web access support, off by default

- BYO API key support for optional web search

- Early backend hooks for future board/task workflows

- Stability and smoke-test improvements

As always, FriedrichAI is still Early Access. Expect rough edges, but the goal remains the same: a practical local AI assistant that runs on your machine, with privacy and user control at the center.

The free demo for FriedrichAI is now available.

The demo is meant to help you check whether FriedrichAI launches and runs on your system before buying the full version. FriedrichAI is a local offline AI app, so hardware compatibility matters.

A few notes:

The demo uses the safest compatibility path, so it may be slower than the full version on supported hardware.

If the demo does not launch or does not run well on your machine, please do not buy the full version yet.

If it does run, the full version adds the main app features, tier options, persistent local memory, and optional media packs.

Thanks to everyone who has tested, wishlisted, purchased, or sent feedback so far. This is still Early Access, and the goal is to keep making FriedrichAI more stable, useful, and easier to run.

Added some spacing and font highlights to make it easier to read what the AI is saying.

A small bug-fix patch is now live for FriedrichAI.

This update fixes an issue where Cyrillic and other non-Latin text could appear blank, broken, or unreadable in the chat input and response area.

What changed:

- Fixed Unicode/Cyrillic rendering in the chat input box.

- Fixed Unicode/Cyrillic rendering in the chat history and AI response area.

- Added a safer fallback for a context-overflow case where recalled local context could cause a failed chat request instead of retrying safely.

FriedrichAI is still English-first for the interface and documentation, but normal chat text should no longer break just because it uses non-Latin characters.

Thank you to the user who reported this and provided screenshots.

FriedrichAI Update — Local Files, Project Tools, and Safer Workspace Support

This update expands FriedrichAI’s local utility tools. The goal is simple: Friedrich should be able to help with real files and real projects without needing cloud access, API keys, or unsafe filesystem permissions.

New: Safe Workspace File Tools

Friedrich can now create and work with files inside its own local workspace folder.

Added commands include:

• /workspace — show the local FriedrichAI workspace

• /savecode filename.py — save the last code block as a real file

• /readcode filename.py — read a saved workspace code file

• /listcode — list saved workspace code files

• /checkpython filename.py — run a local Python syntax check

• /checkjson filename.json — run a local JSON check

• /checkmd filename.md — run a simple Markdown check

• /confirmoverwrite — confirm a pending overwrite

• /canceloverwrite — cancel a pending overwrite

Safety matters here. Friedrich can only write inside its own FriedrichAI workspace. It does not get arbitrary filesystem write access, does not delete files, does not silently overwrite files, and does not allow unsafe paths like absolute paths or ../ traversal.

New: Local Document Converter Lite

Added local document conversion tools:

• /totxt file.txt

• /todocx file.txt

• /topdf file.txt

• /extracttxt file.pdf

These tools are local-only and are meant for simple document conversion and text extraction workflows.

New: Read-Only Project Mode

Friedrich can now scan a selected project folder and help you understand it without modifying it.

Added project tools include:

• /project "C:/path/to/project" — scan a project folder

• /projectfiles — list detected project files

• /projectread "relative/path/to/file" — safely read one project file

• /projectlast — show the last project file read

• /projectfind "SearchText" — search inside the active project

• /projectsearch "SearchText" — alias for project search

• /projectgrep "SearchText" — alias for project search

• /projectmap — create a read-only project flow map

Project mode is read-only. It is designed to help Friedrich understand your project structure, inspect files, follow references, and answer questions with better context.

New: Multi-File Project Context

Friedrich can now keep a small working set of selected project files in context.

Added tools include:

• /projectaddlast — add the last-read project file to the active working set

• /projectctx — show the current project context

• /projectclearctx — clear selected project context

• /projectforgetfile — remove a selected file from context

This helps with questions that need more than one file, such as “how does this system connect?” or “what file should I check next?”

New: Saved Project Shortcuts

You can now save and reopen project locations by name.

Added commands include:

• /projectsave name — save the active project

• /projectopen name — reopen a saved project

• /projectlist — list saved projects

• /projectdelete name — remove a saved project shortcut

This is useful if you work across multiple local projects and do not want to paste the path every time.

Improved: Better Project Follow-Up Grounding

Project follow-up questions now use the active scan, last-read file, and selected project context instead of relying only on chat memory.

That means Friedrich should do a better job answering follow-up questions like:

• “Where is this called?”

• “What file should I read next?”

• “How does this connect?”

• “What changed between these files?”

• “What is likely causing this issue?”

Local First

All of this remains local. No cloud. No API keys. No tracking.

This update does not give Friedrich unrestricted filesystem access. File creation is limited to Friedrich’s own workspace, while project tools are read-only unless a future update adds a separate guarded edit workflow.

This patch improves several core FriedrichAI workflows and cleans up a few rough edges from the latest tool updates.

Changed

• Improved imported file handling so Friedrich answers questions from imported .txt and .md files instead of dumping the raw file back into chat.

• Added /proofread and /spellcheck for quick spelling and grammar cleanup.

• Restored and improved /specs output with clearer hardware/tier information.

• Updated startup guidance with practical command examples for common tools.

• Added backend media availability flags so the interface can tell which optional media packs are installed.

• Updated quick actions: broader Creator/Troubleshoot tools replace narrower Steam Post/Support buttons.

• Improved media button handling for optional ImageGen, VideoGen, and AudioGen packs.

• Fixed duplicate startup “ready” messaging.

Notes Optional media tools still require their matching installed packs. If a pack is missing, the related command will explain that instead of failing silently.

As always, FriedrichAI remains local-first: no cloud account, no API key, and no tracking.

This update improves FriedrichAI’s main interface so tools are easier to find, understand, and use.

The top of the app now shows installed media support at a glance, including Image, Video, Audio, and Voice status. A new Specs button also lets users run a local hardware check to estimate which FriedrichAI tier best fits their current machine.

Quick actions have been cleaned up and reorganized for faster access to common tools like Help, Specs, Doc, Rewrite, Steam Post, Support, Code Check, Code Fix, Image, Video, Creator, and Memory.

Drag-and-drop behavior has also been improved. Code files now route toward code checking, media files prepare the correct media command, and document/text files are imported without assuming what the user wants to do with them.

This is a usability-focused patch. No cloud services, no API keys, and no tracking have been added.

Creator Tools Update — Local Writing Suite Added

This update adds the first version of FriedrichAI’s Creator Tools: a local writing and planning suite built directly into FriedrichAI.

The goal is simple: make FriedrichAI more useful for creators, solo developers, writers, and anyone working from notes, drafts, lore, documentation, or imported project material.

New Creator Tools commands:

/creatorhelp — Show Creator Tools commands and examples. /rewrite — Rewrite pasted text for clarity, tone, or polish. /outline — Turn an idea into a structured outline. /article — Write an article or long-form draft from user-provided notes. /steamupdate — Create Steam-style update posts from patch notes or rough notes. /supportreply — Draft calmer replies to user feedback, bug reports, or support issues. /storecopy — Help write or tighten store page and product copy. /lore — Create lore entries, item descriptions, character notes, or worldbuilding text. /doc — Turn notes into user-facing documentation or guide-style text.

Creator Tools can also work from imported information. You can import notes, drafts, docs, lore, patch notes, or project information, then ask FriedrichAI to write from that material.

Example uses:

• Import patch notes, then use /steamupdate to create a cleaner update post.

• Import lore notes, then use /lore to expand them into character or item entries.

• Import support context, then use /supportreply to draft a response.

• Import project notes, then use /doc or /article to turn them into readable documentation.

FriedrichAI is still fully local. These tools do not require a cloud account, API key, or online service.

Also included:

• Updated startup message to mention Creator Tools.

• Updated /help with Creator Tools and Dev Tools commands.

• Added /creatorhelp for writing-focused examples.

• Added status support for Creator Tools availability.

• Added guardrails so FriedrichAI uses provided/imported material as source context instead of pretending it has live browsing or current external data.

This update continues FriedrichAI’s goal: local AI without the setup maze.

FriedrichAI Dev Tools Update

This patch adds the first major pass of FriedrichAI’s local developer tools.

Friedrich can now check pasted code, inspect code files, summarize file structure, and return full corrected files without using any cloud service or external API.

New commands:

/codecheck — Check pasted or imported code for local syntax/rule issues. /codesummary file="C:/path/file.py" — Summarize a code file. /codefix file="C:/path/file.py" — Return a full corrected version of a code file. /coderules file="C:/path/file.py" — Show the detected language/engine rules.

Supported checks include:

• Python syntax checking using FriedrichAI’s bundled Python runtime.

• Python static checks for duplicate definitions, bad local function calls, mutable defaults, obvious zero division, and empty-container indexing.

• Godot/GDScript 3.x rule checks, including Godot 4 syntax warnings, missing colons, empty function bodies, mixed indentation, and unsafe patterns for Godot 3.5 projects.

• Unity C# rule checks for common MonoBehaviour issues, Update-loop performance mistakes, missing Time.deltaTime movement, and structural problems.

• Unreal C++ rule checks for reflection macros, GENERATED_BODY usage, UFUNCTION formatting, raw UObject/AActor construction warnings, and class-member structure.

• JSON parse checking.

FriedrichAI also now detects obvious pasted code and routes it into local code checking automatically instead of treating it like normal chat.

The goal of this update is not to turn FriedrichAI into a full IDE or autonomous coding agent. The goal is practical local assistance: better checks, clearer warnings, safer file repair, and fewer broken code answers.

As always, FriedrichAI runs locally on your machine. No cloud, no account, no API key, and no data leaving your system.

Also included:

• Updated /help command with Dev Tools commands.

• Shorter startup message.

• Cleaner code-check output.

• Better handling for image-to-video, morph video, 10-second video workflows, and audio command documentation.

• Improved full-file repair formatting so returned code is cleaner and easier to copy.

This is the first Dev Tools pass. More project-aware improvements may come later, but this update keeps the system lightweight, local, and safe.

FriedrichAI Patch Update

A new patch is live for FriedrichAI.

This update focuses on polish, media workflow support, and quality-of-life fixes:

• Added better support for image-to-video and image morph workflows

• Drag-and-drop image handling has been improved

• Dropping two images now prepares a morph command automatically

• File import behavior is cleaner for media files

• Voice output has been changed to off by default

• Voice playback is now treated as optional/experimental

• Fixed mute/voice toggle behavior

• Improved launcher/startup notes for optional media packs

• Export/build cleanup for the current Steam version

Voice output is still fully local, but the current lightweight voice is not where I want it yet, so it will stay optional rather than being pushed as a main feature.

The main focus remains the same: FriedrichAI runs locally on your PC with no cloud account, no API key, and no data leaving your machine.

Thanks to everyone testing the Early Access build and sending feedback. More improvements are coming.

FriedrichAI Video Generation Guide

Image-to-Video (I2V) and Morph-Style Clips

==================================================

1. WHAT IS IMAGE-TO-VIDEO (I2V)?

==================================================

Image-to-video (I2V) starts with an existing image and generates a short video clip from it.

Instead of creating a video entirely from text, FriedrichAI uses your image as the starting point and then applies motion or transformation based on your prompt.

Use I2V when you want to:

- animate a still image

- add movement to a character or scene

- create smoke, cloud, light, or atmosphere effects

- make surreal transitions

- create morph-style clips

==================================================

2. I2V COMMAND FORMAT

==================================================

Use this format:

/i2v "image path" "motion prompt"

Example:

/i2v "your_image_here.png" "slow cinematic motion, smoke drifting across the face, dramatic atmosphere"

Important:

- Put the image path in quotes, especially if it contains spaces

- Put the motion prompt in quotes

- When in doubt, quote both

==================================================

3. HOW TO USE I2V

==================================================

Step 1:

Choose an image you want to animate.

Step 2:

Use the /i2v command.

Step 3:

Provide a motion-focused prompt.

Step 4:

Wait for the clip to generate.

Step 5:

If needed, retry with small prompt changes.

==================================================

4. HOW TO WRITE A GOOD I2V PROMPT

==================================================

For I2V, do not just describe the image.

Describe what should happen in motion.

Weak prompt:

"a woman in a dark room"

Better prompt:

"soft smoke drifts around the subject, hair moves slightly, slow cinematic camera push, atmospheric lighting"

Good prompt structure:

subject motion + visual motion + transformation or atmosphere + style

Examples:

- "blue smoke swirls around the subject's face, slow cinematic motion, dramatic lighting"

- "clouds drift behind the character, subtle movement, surreal dreamlike atmosphere"

- "the subject remains still while fog rolls through the frame, slow moody camera movement"

Helpful motion words:

- swirls

- overtakes

- wraps around

- expands

- dissolves

- reforms

- reveals

- transforms

- drifts

- clears

- pulses

- rises

- spreads

==================================================

5. WHAT IS A MORPH-STYLE CLIP?

==================================================

A morph-style clip is a short transformation video.

This usually means one thing visually changes into another over a few seconds.

Examples:

- smoke covers a face and reveals a different face

- a figure dissolves into fog and reforms

- clouds overtake the head and clear to reveal a new person

- a robotic face shifts into a human face

Morph-style clips work best when treated as one short transformation, not a whole movie scene.

==================================================

6. MORPH COMMAND FORMAT

==================================================

Morph clips use the same I2V command:

/i2v "image path" "motion prompt"

Example:

/i2v "your_image_here.png" "thick smoke slowly overtakes the subject's head, the silhouette reshapes, then the smoke clears to reveal a different person, surreal cinematic transformation"

==================================================

7. HOW TO WRITE A GOOD MORPH PROMPT

==================================================

A good morph prompt usually includes:

1. the starting subject

2. what overtakes or transforms it

3. how the transformation happens

4. what is revealed at the end

5. the visual style

Best prompt structure:

subject + transformation + motion + reveal + style

Example formulas:

- "smoke rolls over the face, the features dissolve, then the smoke clears to reveal a different person"

- "clouds wrap around the head, the silhouette shifts, then the clouds break apart to reveal a new character"

- "black mist consumes the subject, reforms into a new shape, and clears to reveal a different face"

==================================================

8. GENERIC I2V EXAMPLES

==================================================

Example 1:

/i2v "your_image_here.png" "soft fog drifts through the scene, subtle cinematic motion, moody atmosphere"

Example 2:

/i2v "your_image_here.png" "blue smoke curls around the subject, hair moves slightly, slow surreal motion"

Example 3:

/i2v "your_image_here.png" "clouds move behind the figure, gentle movement, dreamlike lighting"

Example 4:

/i2v "your_image_here.png" "the camera slowly pushes in while light flickers across the scene, dramatic atmosphere"

==================================================

9. GENERIC MORPH EXAMPLES

==================================================

Example 1:

/i2v "your_image_here.png" "smoke slowly overtakes the subject's face, the head dissolves into cloud, then the smoke clears to reveal a different person, surreal cinematic transformation"

Example 2:

/i2v "your_image_here.png" "dark mist wraps around the figure, the silhouette reshapes, then the mist dissolves to reveal a new face"

Example 3:

/i2v "your_image_here.png" "clouds cover the subject's head and shoulders, distort the outline, then break apart to reveal a different character in the same pose"

Example 4:

/i2v "your_image_here.png" "a glowing vapor overtakes the face, reforms into a different structure, and clears to reveal a transformed version of the subject"

==================================================

10. TIPS FOR BETTER RESULTS

==================================================

- Use a clear starting image

- Keep the subject centered when possible

- Ask for one main action or transformation per clip

- Keep the motion simple and focused

- Think in short 5 to 10 second scenes

- Retry with small prompt changes

- Save the best clips and build longer sequences scene by scene

Best approach:

Do not try to generate a whole movie moment at once.

Generate one strong shot at a time.

==================================================

11. COMMON MISTAKES

==================================================

Mistake:

Only describing the image.

Better:

Describe the motion or transformation.

Mistake:

Asking for too many things at once.

Better:

Focus on one main visual event.

Mistake:

Using a weak or unclear starting image.

Better:

Use a clean image with a clear subject.

Mistake:

Expecting perfect results on the first try.

Better:

Treat it as experimental and iterate.

==================================================

12. QUICK REFERENCE

==================================================

Basic I2V:

/i2v "image path" "motion prompt"

Basic Morph:

/i2v "image path" "transformation prompt"

Simple I2V example:

/i2v "your_image_here.png" "soft smoke drifts around the subject, cinematic motion"

Simple Morph example:

/i2v "your_image_here.png" "smoke overtakes the face, reshapes the silhouette, then clears to reveal a different person"

==================================================

13. FINAL NOTE

==================================================

Image-to-video and morph-style generation are experimental features.

Best results usually come from:

- clear starting images

- strong motion prompts

- short scenes

- multiple attempts

- small prompt refinements

If you get a good result, save it and build from there.

That is the best way to use local video generation right now.

FriedrichAI is on sale now, and the free optional media generation DLC packs are live.

You can now add local text-to-image, video generation, and audio generation support to FriedrichAI through free DLC.

No cloud. No API keys. No accounts. No tracking.

FriedrichAI runs on your machine, using your hardware, with optional local media tools you choose to install.

This is still Early Access, and the media tools are experimental, but the goal is simple:

Local AI should not require command-line punishment.

Your AI. Your machine.

Hey guys, some pretty big backend work being done today. Image gen will be down for a little bit while I clean up pathing. That will be fixed hopefully by tomorrow but there is good new. With that come the ability to generate videos and audio resources depending of course. If there are any questions or concerns, bugs please reach out in the forums. Thank You all

Your Humble Dev

Hey everyone, I am sincerely sorry about the need to do this but a bug has been discovered that needs to be addressed. The memory wipe was too weak before. I have hardened it but unfortunately that fix will reset your progress. I hope this doesn't affect anyone too terribly. I hate doing this and promise to do my best not to have to ever again.