Daily · Jul 15, 2026

Higgsfield funds filmmakers; Pika edits footage by prompt; DeepMind turns video into 4D

Distribution and interface moved together: Higgsfield opened a large filmmaker grant, Pika made footage editable by prompt via Gemini Omni on its MCP, and DeepMind's GenCeption turned video into depth, segmentation and searchable 4D.

Industry01

Higgsfield opens a 100,000-credit Filmmaker Grant

Higgsfield launched a Filmmaker Grant of 100,000-plus credits with uncapped slots, open to directors, artists and first-time users, judged on a reel by July 29. Uncapped, talent-only funding is a bid to become the default place AI filmmakers start, ahead of rivals courting the same creators. Treat it as near-free production budget: put a real project through it, but keep deliverables and source files portable so the work survives outside one platform's credit economy.

Source Higgsfield AI @higgsfield_ai · 2026.07.15

Tool02

Pika turns footage into 'anything' by prompt via Gemini Omni

Pika wired Gemini Omni into the Pika MCP so existing footage can be transformed by prompt, pushing post-production onto the same conversational surface as generation. When relight, reframe and swap run behind a single instruction, the studio's edge shifts from owning an editor to owning the judgment of which prompt to run. Use it as a fast variant-and-localization lever after continuity is locked, and keep a human gate on brand-critical calls a prompt can't be trusted to make.

Source Pika @pika_labs · 2026.07.15

Model03

DeepMind's GenCeption turns video into searchable 4D

Google DeepMind's GenCeption turns ordinary video into depth, segmentation, 3D keypoints and searchable 4D worlds. For visual teams this points past flat generation toward spatially aware footage you can query and re-stage, which is the substrate for consistent relighting, reframing and object edits. Watch it as the layer that could make continuity and 3D-consistent edits reliable, and start noting which of your recurring tasks (camera moves, object removal, relight) would benefit from a real depth/segmentation pass rather than a 2D guess.

Source Min Choi @minchoi · 2026.07.15