Last Week’s Big Moves in Image & Video Generation
Rapid change hit AI media models last week. OpenAI drew closer to deploying its next-gen image model, GPT Image 2, and Shutterstock unveiled a commercial‑ready AI video generator. Here’s what solidified the future.
OpenAI Nears Launch of GPT Image 2
OpenAI looks set to retire DALL‑E 2 and DALL‑E 3 by May 12, 2026, replacing them with its upcoming GPT Image 2. The model surfaced in early testing—via LM Arena and Reddit—showing off near‑perfect text rendering and native 4K output in a single‑pass generation process, a major leap over prior pipelines. While not yet confirmed, analysts expect a late April to mid‑May release window.
This shift hints at faster, cleaner, more realistic image generation that could redefine visual content workflows.
Shutterstock Rolls Out AI Video Generator
Shutterstock launched its AI Video Generator, integrating powerful text‑ and image‑to‑video models with its vast licensed media library. Designed for enterprise and creative professionals, the tool supports ideation through to go‑to‑market execution, with clear commercial usage rights baked in. A significant turn toward enterprise‑grade, generative video tools.
Recap: Q1 2026 Model Highlights
ByteDance Seedance 2.0 Surges Ahead
Seedance 2.0 remains the top performer on AI video model benchmarks, crossing ahead in both text‑to‑video and image‑to‑video scoring. Its strengths include native audio‑visual synthesis in one pass, phoneme‑level lip sync, multi‑shot continuity, and multi‑reference capability. It’s the go‑to choice for narrative and action‑driven content.
Kling 3.0, Veo 3.1 & Runway Gen‑4.5: Specialists Shine
- Kling 3.0 delivers native 4K output, multi‑shot storyboarding, and multilingual synchronized audio—great for serialized short ads or narrative clips.
- Veo 3.1 brings a friendly free‑tier model into Google’s Vids, targeting casual creators with up to ten free generations per month.
- Runway Gen‑4.5 offers robust scene control with tools for keyframes, layout, pacing, and consistent temporal quality—perfect for polished, film‑style edits.
What It Means for Creators and Teams
1. Faster, Richer Image Generation
GPT Image 2’s single‑pass, high‑resolution generation could set a new baseline for visual content. Expect reduced rendering times, cleaner visuals, and better integration with generative chat workflows.
2. Ready‑Made Video Content with Licensing Peace of Mind
Shutterstock’s solution closes the gap between creativity and compliance. Teams can now generate polished video assets backed by licensing clarity—no more guesswork on rights.
3. Choose by Use Case, Not Just Hype
Seedance 2.0 dominates benchmarks across the board. But Kling 3.0 brings multi‑shot storytelling, Veo 3.1 lowers the barrier for casual creators, and Runway Gen‑4.5 offers editorial precision. Pick the model based on your workflow, not just buzz.
Bottom Line
Last week marked a turning point. OpenAI prepares a major upgrade with GPT Image 2. Shutterstock pushes video generation into enterprise workflows. And Seedance 2.0 holds the lead in model quality while niche players fill key workflow gaps.
Try multimodal chat from all providers, image‑generation models, and hybrid RAG on your own data in dedicated workspaces with Projectchat.ai – start your trial now at https://projectchat.ai/trial/
