The Render

Google moved fast to fill the vacuum left by Sora's exit last week, dropping Veo 3.1 Lite — a significantly cheaper tier of its video generation platform — and cutting prices on its existing Fast tier at the same time, leaving itself as essentially the only Western player with a serious AI video offering at scale. For production and post shops exploring AI video as a practical tool rather than a novelty, the economics just got a lot more approachable.

Meanwhile, ByteDance quietly embedded Seedance 2.0 inside CapCut's editing timeline, and Suno dropped what may be its most consequential update yet, letting you train the model on your own voice and catalog.

What happened: Google introduced Veo 3.1 Lite, its most cost-effective video model, delivering the same generation speed as Veo 3.1 Fast at less than half the cost. The model supports Text-to-Video and Image-to-Video, 720p or 1080p resolutions, landscape and portrait aspect ratios, and is available now via the Gemini API and Google AI Studio. A further price cut on the Fast tier arrives April 7.

Why Is This Important? With Sora gone and ByteDance's Seedance global rollout stalled by IP concerns, Google is now the dominant Western platform for AI video generation — and this week it got substantially more affordable. For production and post shops that have been watching from the sidelines, waiting for the price-to-quality ratio to make sense, that conversation just shifted. Pre-viz, B-roll generation, ad creative iteration — the use cases that were hard to justify six months ago are worth revisiting now.

TechCrunch

What happened: ByteDance confirmed that its Dreamina Seedance 2.0 model is now rolling out in CapCut, allowing creators to draft, edit, and sync video and audio content using prompts, images, or reference videos. The phased rollout begins in Brazil, Indonesia, Malaysia, Mexico, the Philippines, Thailand, and Vietnam — notably not the U.S., while ByteDance addresses IP protection concerns flagged by Hollywood. The model supports clips up to 15 seconds across six aspect ratios.

Why Is This Important? The strategic play here isn't the model — it's the address. Generation living natively inside CapCut's edit timeline means it becomes a production primitive rather than a separate tool. In CapCut, the model rolls out across editing features such as AI Video and generation tools like Video Studio. For commercial shops already delivering social content through CapCut, this compresses the workflow significantly. The U.S. holdback is worth monitoring — it signals the IP battles are far from settled.

Note: remember to read those “terms of service” folks…

The Verge | Suno Blog

What happened: Suno unveiled v5.5, emphasizing user control through three new capabilities: Voices, My Taste, and Custom Models. Voices — the most-requested feature — enables users to train the vocal model using their own voice, submitting clean acapellas, fully produced tracks, or a phone recording. Custom Models lets Pro and Premier subscribers train Suno on their own catalog to generate music aligned with a specific style. My Taste adapts over time to a user's genre and mood preferences.

Why Is This Important? For post houses using Suno for temp tracks and lower-budget scores, Custom Models is the meaningful shift — it's the difference between generic AI music and something that can approximate a specific palette or established sonic identity. The Voices feature raises immediate questions for sound designers and music supervisors: at what point does this become a credible sketch tool for topline development, or a production resource for demo-quality guide vocals? Voices and Custom Models are Pro and Premier tiers only.

PetaPixel

What happened: Runway showed off new technology at NVIDIA GTC where HD video generates in real time, with time-to-first-frame under 100 milliseconds, running on NVIDIA's next-generation Vera Rubin architecture. The as-yet-unnamed model works closer to how a game engine renders frames, streaming them continuously rather than generating a finished clip after a prompt. It's currently a research preview — not shipping yet.

Why Is This Important? This is a different category of capability than anything currently in production workflows. Just a few months ago scenes were generated in minutes, now it's seconds — and now Runway is demonstrating sub-100ms. The near-term implications for interactive pre-viz, real-time virtual production environments, and live broadcast are significant. The longer-term implication — AI characters and worlds that respond live to whatever is happening around them — is the one worth sitting with.

PRNewswire

China's iQIYI launched Nadou Pro on March 30 — the country's first AI agent built specifically for professional film and TV production, supporting end-to-end workflows from script development and storyboarding to final output. Thailand Business News A director credited the tool for achieving the mecha aesthetic in a film that premiered the same week.

Western coverage is thin, but the scope is ambitious — if integrated AI production agents prove out commercially in Chinese productions first, those patterns will (are likely to) migrate quickly.