Use the menu in the top left to select what tool you want.

Your Account

Notices

6-2 System restored. Gokstad AI was shut down for a little bit to test new hardware. The new hardware opens the possability for AI redundancy. Once that is accomplished, shutdowns like these will no longer obstruct service.
5-24 Mimic updated. Mimic is now fully functioning with a fast/slow mode. And it now has a progress report in the output window that also keeps your browser form timing out and throwing a false error dueing a generation. I will be returning to Mimic to make more improvements but for now, I'm going to celebrate.
5-24 Mimic updates. Mimic is not fully functional right now. I tried updating it to the model Rúna uses for her voice. This cut generation time by 90%, but it lost fidelity to the source voice's mannerism and style. So I reverted it back and added a toggle switch to allow you to choose which mode you want to use. One gives speed but you sacrifice realism. The other gives you very good realism but you sacrifice speed and it sometimes fails on certain situation. This is my weekend project and when I'm done, Mimic will be better than before.
5-21 Mimic issue. On generation over 70 words, Mimic is throwing an "Error" message. However this is likely a false error caused by browser timeout. I will be modifying the UI soon to have the output streamed as it's being generated. This will allow you to start getting generation outputs as it's being made instead of waiting the whole generation to complete.

5-17 Developer's Button in Rúna. I added a developer's notes button to Rúna's interface that will itterate her updates and planned improvements that I post here in a move to increase transparancy. And right now I will be adding a microphone button as well that will use a forked version of Scribe to allow voice-to-text. This will be a phase 1 to making her a total talk-to-talk bot.
5-11 Naughty Rúna. An audit of interactions for hallucinations revealed her “Self knowledge” system is failing. This is a what instructs her on what she is, what Gokstad is, what her capabilities are, and what she can do. Without this, like all models, she will tell users she can do whatever they ask her to do even when she can’t. Currently all she can do is perform searches and help with basic code. I will take some time to harden this up and get her to provide links to Gokstad’s other tools like Skyr and Hullwave instead of claiming she can do their functions herself.
This week I have been working hard to create a system that will give her the ability to learn and recall actual facts in real time, but this is a complex system and will take a long time.
4-30 Rúna's copy button. I finally got around to adding a copy button to Rúna so you can now copy her responses. Better late than never.
4-27 Skyr update. Models and LoRAs are no longer locked together. You can now mix-and-match them as you like.
4-20 Hullwave Upgraded. Hullwave's music generation model has been upgraded to a bett version.
4-13 Mimic Upgraded. Voice sample optimization is now fully automated. And it is now using the same .npz file type and per-processing that made Rúna's voice so fast and more dynamic. Before Mimic would have to process every voice sample as part of a generation.
4-12 Mimic Upgrading. I am currently upgrding Mimic by automating the optimizer so it is performed immediately when a file is uploaded. I also am taking the advancements accomplished with Rúna's voice to improve Mimic's performance. At this time Mimic may have issues during development.
3-20 Inturruption of Service March 27. The internet provider has alerted me that they will be performing maintenance from Midnight March 26th until 1 AM March 27th. During this time, Gokstad and Dokstad AI will not be funcitoning.
3-13 Rúna now has response feedback. I gave her responses a thumbs up/down system that tags the responses. This is to help improve her in the future. Liked responses will be added to reinforcement datasets and poor or wrong responses will be added to corrective datasets.

3-12 Rúna Can code. Rúna is now a multi-model chatbot. If you talk about coding with her, she shifts from her chat model to a model specially trained for code. I added code windows with copy icons. And I set up her voice so she handles the code blocks instead of just reading them.
3-7 Introducing Arcader. I spent the last couple days combining Skyr and the development I did on Reflekt into a new tool for Game developers. Instroducing ARCADER. This is a tool that provides a list of vector srite sequences and combining them with tghe user's prompts, creates complete sprite animations. Even is the individual images aren't perfect, the amount of time this tool will save indi-game devs is priceless. I expect to have it ready to go by the middle of next week.
2-27 Visionary is alive. The Gokstad video generator Visionary is now successfully generating video. The quality of the output is not impressive if you don't understand the hardware limitations it is running on. This system can't operate across multiple GPUs, so Gokstad needs BIGGER GPUs, not more of them. A GPU big enough to generate quality video is next on the shopping list. Your support helps make this become a reality.
2-24 Rúna's voice has been vastly improved. Now instead of waiting for you to press the speaker button to hear her read aloud, the voice is automatically generated and already being built by the time you click the button. There is still room for more improvement and I will be tackling these issues. One such is the audio is not compressed enough so it may buffer on mobile devices, but I know the issue and I'll tackle it shortly.
2-17 Rúna now includes her sources. If she does a search, there is now a "Searching" spinner and then a "Sources" button that will appear. This is the information her search system found. Currently her searcher is poor, BUT it's finctioning.
2-16 Rúna now has a RAG function and performs searches. I just installed a system originally developed for Gokstad's Tyr search RAG into Rúna. Certain questions trigger a search function, but this is still very new. Her information is only as good as what she gets in return.
2-12 Skyr welcomes Amelia and a slew of new styles. You can now generate Amelia by name. Just type "Amelia ___" and it will create her. I have also been adding new styles like a mad-man, as well as improving the style guides to produce better results instead of just relying on user's prompts.
2-6 Rúna is no longer a base model. Rúna has officially had her first training session and is no longer a base model, she is officially "Rúna-v1.0". Her responses will now be slightly different. As to the full effect of the training, it is yet to be determined.
2-5 Resources will be limited 2-5 - 2-14. I am running a massive training session that requires all the GPU I can spare. Tyr has been shut down and Runa's voice has been shifted to CPU, which makes it very slow but still active.
2-4 Mimic is restored. Mimic is back up and abiding by the GPU assignments.
2-4 Mimic is down. While testing the new GPU I discovered Mimic wasn't respecting the GPU assignment system and just picking one for itself which will conflict with other apps. It's down as I work on fixing this issue.
2-4 System will be shut down for 30 minutes. I'm temporarily shutting the system down for a bit to install a new GPU. Eventually Tyr will have a brother server so when I have to shut down for an upgrade, there will be a redundant system to carry the load.
2-2 Skyr improved and new models. I fixed Skyr's LoRAs and added PonyRealism V23, Synthwave V1 & V2, and DSLR Portrait to the list.
1-27 Scribe is back up. Scribe is not only back up, but now it supports images too. So if you need an image posted to the internet transcribed to actual text, Scribe can now do it.
1-26 Scribe is down for now. I'm updating Scribe so it will accept images and PDFs as well. It's currently broke and it's late. I'll get on it when I get up.
1-24 Mimic is fixed. I updated it and it's working again. It's still very slow however. I'll be working on improving this.
1-24 I'm fixing Mimic. Mimic's system is outdated causing it to fail. I am updating it now.
1-16 The AI systems are back up. The training session has completed. There will be more in the future and as Gokstad grows, shutting the system down for training sessions will eventually be a thing of the past.
1-15 The AI is currently down for training I shut off the AI systems for the next 24 hours to allocate resources to a training session. Tuberipper is still operational.
1-8 TubeRipper. I added TubeRipper, a Youtube downloader for Supporter tier and above. It's not AI driven but it is a valuable tool for many people and personally I myself got tired of using unsafe sites to do it.
1-2 Hullwave is back online. I spent all last night and today working on it. It's back online with an improved backend, but there is still an issue loading audio into mobile devices.
12-30 What's going on/updates. I have been spending the last week developing a coding workstation called Bruno. It is one of the most complex user interfaces I have ever built. Each new function added to it creates new challenges and issues that need to be fixed. The system that handles all file transfers for Gokstad AI had almost trippled in size due to this one app. Up until now the singular system has been fine, but just now I discovered an single inproper indentation with what I was working on caused three other apps to lose their file management. I fixed this issue immediately but this brings about a new question of breaking the system up so each app has its own system. This is to to simplify it into smaller systems as well as make each app more secure, BUT this would be a start of a major overhaul.
12-24 Hullwave's library issue? Hullwave had an issue not loading some of the user's library files on mobile devices. But it mysteriously is not doing it now. This may have just been my mobile device used for testing. If you experiance this yourself, reach out to me on our sociel media link.
12-24 Scribe is back up. It was a coding battle, but Scribe is back up and running better than before.
12-23 Scribe is currently down. Scribe is getting an error on transcriptions past the first one. I'm working on fixing this right now.
12-20 Hullwave's library issue. Hullwave has had an issue not always pulling up the user's library files on mobile devices. I will be working to fix this soon.
12-20 Rúna is back up. She is now back up and using her new system and as a bonus, she can now comprehend time. Her memory seems to be working but I will be monitoriing the rest of the night to make sure it's working right.
The issue that was effecting Rúna exists on all the apps. I will be going through them one-by-one over the next few days and updating them as well.
12-19 I found the issue. The problem was in how the systems police themselves. When there's an issue, the system fails/locks up and is unable to deal with the issue because it's all tied together. So I'm redesigning the systems to be seperate from their generation functions. This way if a generation fails/locks up for some reason, the system itself isn't locked up, detects the problem and can attempt to deal with it.
12-19 A common bug is killing generation. I discovered that it wasn't just Rúna having an issue, it's most of the apps. There is a process that is randomly causing restarts mid-generation. I'm trying to identify the cause so I can fix it.
12-16 Audio Optimizer fixed. A small change to Mimic was the cause to the optimizer failure. I fixed it.
12-14 Hullwave added to menu. Hullwave, our text-to-music system has been added to the menu. It works but has a couple minor bugs I'm working on.
12-13 The audio optimizer is off. I thought I had fixed the issue, but it broke again. After I launch Hullwave I will get on this.
12-10 Rúna has an issue where the second interaction with her, may crash. This is on the To-Fix list.
12-7 Scribe is partially working. I discovered that Scribe was having some issues staying on the assigned GPU during long transcriptions.
11-15 Scribe is now live. Transcribe speech from audio and video into text.
11-14 Rúna memory upgrade. Conversation context and recall are now much more stable and lucid.
11-11 Mimic model improved. Better phonetic look-ahead for more natural speech.
11-10 Mimic optimizer added. Automatically picks the best clip from your upload and cleans it for cloning.
10-21 New tools in development. Reflekt (image style transfer), Plagiarist (text re-authoring), and a video generator pipeline.

Membership

Current tier: FREE

See membership plans

Gokstad AI is funded directly by its users. Paid tiers unlock higher usage for Skýr, Mimic, Scribe, and future tools, and help keep Tyr’s GPUs fed and cooled.

System

Your contributions improve our systems.

Account Settings

Change Email

Current Email: Loading...