v1.8.0

gristlabs/grist-corev1.8.0Feb 22, 2026by FranckyB

AI Summary

Introduces a Prompt Hub for managing prompts across tools, adds Ollama as an LLM backend, and features 'Split by Paragraph' for batch audio generation.

Key Highlights

  • Built-in Prompt Hub with category filtering
  • Ollama support as a new LLM backend
  • Split by Paragraph feature for batch audio generation

New Features

  • One-click prompt transfer to generation tools
  • Ollama model selection in settings
  • Batch audio generation with name collision checks

Full Release Notes

#### Version 1.8.0 - Prompt Hub, Ollama Support & Split by Paragraph

**Prompt Hub Integration**
- **Prompts Directly in Tools** - Every generation tool (Voice Clone, Voice Presets, Voice Design, Conversation, Sound Effects) now has a built-in Prompt Loader accordion for quick access to saved prompts
- **One-Click Transfer** - Select a saved prompt and send it straight to the tool's text input without switching tabs
- **Category Filtering** - Prompts are organized by category, with each tool showing only the relevant prompt type
- **Auto-Refresh on Expand** - Prompt list refreshes automatically when the accordion is opened, no manual refresh needed

**Ollama Support**
- **New LLM Backend** - Added Ollama as an alternative LLM backend for prompt generation alongside llama.cpp
- **Configurable in Settings** - Choose between llama.cpp and Ollama in Preferences; Ollama connects to a locally running instance
- **Any Ollama Model** - Use any model available in your local Ollama installation for prompt generation

**Split by Paragraph**
- **Batch Audio Generation** - New "Split by Paragraph" checkbox in Voice Clone splits text by line breaks and generates a separate audio clip for each paragraph
- **Base Name Prompt** - Enter a base name for the batch; clips are numbered sequentially (e.g., MyScene_001, MyScene_002)
- **Combined Preview** - A combined preview of all clips plays automatically in the audio player after generation
- **Name Collision Check** - Client-side validation warns if files with the same base name already exist in the output folder
- **Works with All Engines** - Compatible with Qwen, VibeVoice, LuxTTS, and Chatterbox