Voxta 140 Update: New Memory, More Stability and Faster Performance.
Added 2025-02-17 21:10:22 +0000 UTCHey everyone!
We just released Voxta v140, and it brings some major improvements! Here’s what’s new:
New Memory System
Memory has been one of the biggest challenges for Voxta, but with this update, we've made a huge leap forward. Voxta now uses Microsoft Semantic Kernel, replacing the old system with a more advanced and stable memory model.
No more keywords—just create an entry, add details, and let the memory system handle the rest.
It uses vector search with cosine similarity, meaning it can recognize similar concepts and retrieve them in real time while you're chatting.
Want to speed things up? You can ask a large language model to generate an entire memory document for your character, drop it into Voxta, hit Extract Memories from Document, and it’ll handle everything automatically.
More Stability & Performance Improvements
New database: We’ve replaced LiteDB with SQLite for a faster and more reliable experience.
Optimized memory prompting: No more keyword-based lookups—just natural memory retrieval.
Chain of Thought improvements: Messages now expire over time, plus a new manual mode (/trigger cot).
Better Conversations & New Features
Multi-paragraph responses: Voxta can now read multiple lines instead of stopping at the first.
Default narrator: You can now assign a narrator character for automatic story narration.
More wake words: Expanded Azure Wake Words support for better voice activation.
Increased context window: Now set to 4096 tokens by default for better long-form conversations.
Fixes & Other Improvements
Updated LlamaSharp (0.21.0) and Coqui (DeepSpeed 0.16.3).
Improved memory search: Voxta now looks up entire words in memory instead of just keywords.
FFmpeg & FFprobe auto-setup: F5 installs now include these tools automatically.
Better transcription timing: Added a 300ms delay after speech starts to prevent cutting off responses.
UI fixes: Better memory book editing, flag expiration info, and more.
This version brings a smoother, faster, and more intelligent Voxta experience. Thanks to everyone for your feedback and support!
How to install Voxta server app: https://youtu.be/1I9VkJ8tTlo
How to update Voxta server app: https://youtu.be/5aa7sducwoc
Comments
I cant get this dumb thing to work so im unsubscribing.
Cloviz
2025-07-23 18:51:06 +0000 UTCWhere do you download the latest version?
Already Replaced
2025-04-11 18:51:12 +0000 UTCAre you adding support for Ollama?
Mark Phillips
2025-03-27 19:17:59 +0000 UTCHello. I want to use the VOXTA-Cloud with the VAM voxta plugin. What IP-Adress/DNS do I need? The API Key I put in the plugin already. Thank you
aprisma
2025-03-03 16:52:56 +0000 UTCI just did something incredible with Voxta. You can combine PhonerLite(free)+Virtual Audio Cable (35€) with your Home Phone. And basically call your Voxta LLM at your home and chat with it over landline phone with no INTERNET whatsoever.
Jeeebus777
2025-02-28 15:35:31 +0000 UTCJe fais que payer votre application elle marche pas qu'est-ce qu'il faut faire pour récupérer des crédits
christophe pieralli
2025-02-25 20:20:16 +0000 UTCIt would be great to add a few lines above how to update - e.g. update as always - see here or: new update method, see here or if you update as always don't forget to install these new serveses
Pullerman
2025-02-20 10:00:31 +0000 UTCThat's exactly why we provide fully local services for you, where you don’t need to pay anything.
Voxta
2025-02-18 17:55:19 +0000 UTCI love the Voxta-Service Software, but it's realy too expensive. I hate the Coin-System. I would love the idea to switch to a mothly "unlimited" and cheaper system.
Purtox
2025-02-18 12:44:58 +0000 UTCThe Text Generation Web UI service is not working. This will be fixed in the next hotfix. We apologize for the inconvenience. In the meantime, you can use local alternatives such as ExLlamaV2, Llama.cpp, or KoboldAI services.
Voxta
2025-02-18 10:34:55 +0000 UTCFWIW - my setup is running everything locally, hitting webui located on a separate PC so an 'old' 3090 handles all the LLM work.
Smumblepooch
2025-02-18 05:23:23 +0000 UTCThe failed install of version 140 actually caused my older versions to stop working as well... "The response ended prematurely. (ResponseEnded)"
WaywardProphet
2025-02-18 01:36:47 +0000 UTCSame issue.
WaywardProphet
2025-02-18 01:31:55 +0000 UTCGot this when doing the initial setup after selecting 'Text Generation Web UI' and clicking 'Next' in the 'Setup Text Generation' screen: Failed to install service: The instance of entity type 'Preset' cannot be tracked because another instance with the same key value for {'UserId', 'LocalId'} is already being tracked. When attaching existing entities, ensure that only one entity instance with a given key value is attached. Consider using 'DbContextOptionsBuilder.EnableSensitiveDataLogging' to see the conflicting key values.
Smumblepooch
2025-02-17 23:32:59 +0000 UTC
