14 / Mar 31 – Apr 6, 2025 · 9 RELEASES

AI News Week 14, 2025

Week 14 2025 AI releases.

Language Models (1)

Dolphin Gemma
Google · Text Generation
Dolphin Gemma is a fine-tuned version of Google's Gemma language model optimized for instruction-following and dialogue tasks. It demonstrates improved performance on conversational benchmarks through specialized training techniques.

OmniSVG
Fudan University & StepFun · Text-to-Image
A unified framework for end-to-end multimodal SVG generation that tokenizes SVG commands into discrete tokens and leverages pre-trained Vision-Language Models to generate complex vector graphics across multiple modalities (Text-to-SVG, Image-to-SVG, Character-Reference SVG). Introduces MMSVG-2M, a 2M-asset multimodal dataset with standardized evaluation protocols for SVG synthesis tasks.
InstantCharacter
Tencent Hunyuan & InstantX Team · Text-to-Image
A diffusion transformer framework for scalable character customization that achieves open-domain personalization across diverse appearances while maintaining high-fidelity results and text controllability, trained on a 10-million-sample dataset with paired multi-view and unpaired text-image combinations.

ACTalker
HKUST, Tencent, Tsinghua University · Text-to-Video
A video diffusion framework for talking head generation that supports multi-signal control (audio, facial motion, or both) using parallel mamba structures with masked selective state space modeling to coordinate temporal and spatial coherence without control conflicts.
OmniTalker
Alibaba Group (Tongyi Lab) · Text-to-Video
End-to-end unified framework for real-time text-to-speech and talking head video generation with synchronized audio-visual output and zero-shot style replication. Achieves 25 FPS inference using dual-branch diffusion transformer with cross-modal fusion.
UniAnimate-DiT
Ali-ViLab · Text-to-Video
A diffusion transformer framework for unified animation generation that enables text-to-video and image-to-video synthesis with improved temporal consistency and motion control capabilities.

Any-Agent
Mozilla AI · Agent Framework
An open-source agent framework that enables building autonomous agents with flexible tool integration and multi-step reasoning capabilities.
ComfyUI-Griptape
Griptape · Agent Framework
Integration plugin that connects Griptape's AI agent framework with ComfyUI, enabling visual node-based workflow creation for building and orchestrating multi-step AI agent tasks.

ComfyUI-UNO
jax-explorer · AI Infrastructure
Extension for ComfyUI that appears to add unified node operations (UNO) functionality, likely streamlining workflow composition and node management in the ComfyUI node-based UI framework.

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.