AI Film Weekly — control room with monitors

Three stories own this week. ByteDance's Seedance 2.5 went to full capacity across the OiiOii platform. OpenAI pinned a date on Sora's API coffin. And Kling 3.0's AI Director started storyboarding your scenes for you — six shots in one pass, no editing timeline required.

Here's what matters and what to do about it.

Seedance 2.5 opens the floodgates

ByteDance launched Seedance 2.5 globally on Dreamina on July 31, and on August 7 OiiOii AI announced the model is live at full capacity — no queue. The specs that matter for filmmakers:

• Up to 50 reference assets (images, videos, audio) in a single generation
• Clips up to 30 seconds — double the old ceiling
• Targeted editing — change one part of a scene without regenerating the whole clip
• From $0.097 per second

That's the stitching problem dying in real time. At 30 seconds per clip with 50 reference inputs, you're no longer assembling a movie from 6-second scraps. You're directing scenes.

Sora's countdown: September 24

OpenAI discontinued the Sora web and app experiences on April 26. Now the API has a date too: September 24, 2026. No video generation is planned inside ChatGPT. If Sora is in your production pipeline, you have roughly six weeks to migrate.

The landing spots: Veo 3.1 for photorealism, Kling 3.0 for cinematic motion on a budget, Seedance 2.5 for multimodal control and native lip-sync. Don't wait for the shutdown — test your existing Sora prompts against all three this month and pick your keeper.

Kling 3.0's AI Director changes the unit of work

Kuaishou's Kling 3.0 keeps stacking features that sound like a film school syllabus. The AI Director generates up to six distinct shots in one pass — wide to close-up, with cinematic transitions handled automatically. Native audio syncs sound and dialogue, with lip-sync in five languages. Elements 3.0 locks character consistency using video references instead of static images.

The practical shift: the unit of work in Kling is no longer a clip. It's a scene. That's what "everyone can be a filmmaker" was supposed to mean.

Google Omni and the realism arms race

Google's Omni generation is now producing clips that viewers can't distinguish from real footage. For independent filmmakers that's a gift — for studios and platforms it's a watermark arms race that's only beginning. Expect disclosure tooling (C2PA-style tagging) to become a production requirement, not a nicety, before the end of the year.

"The model didn't get cheaper. The stack got smarter."

— The 2026 AI Video Production Playbook

This week's reading:
OiiOii AI launches Seedance 2.5 at full capacity | Dreamina launches Seedance 2.5 | The 2026 generator comparison | Kling's AI Director breakdown

GenFlick is building the AI filmmaking stack.

Join 2,000+ filmmakers and creators who read this newsletter every week.

Sign Up Free →