What problem does it solve? AI-generated video clips drift in character identity, hands, and physics beyond a few seconds, and raw clips cannot play in Summer Engine without transcoding and scene wiring. This Skill produces consistent multi-shot cutscenes by locking the look with a reference image, chaining 5-10 second image-to-video shots, and importing them as Ogg Theora into a fading VideoStreamPlayer controller scene. ## Core Features & Use Cases - Reference-Locked Generation: Generate a still with summer_generate_image first, then pass it as imageUrl to every summer_generate_video call so the character looks identical across shots. - Model and Cost Routing: Choose between ltx ($0.10 iteration), kling ($0.50 hero shots), minimax ($0.25 stylized), and veo3 (text-to-video only, synced lip motion), with async job polling via summer_check_job. - Engine Integration: Transcode MP4 to Ogg Theora with ffmpeg, import via summer_import_from_url, and build a cinematics scene with ColorRect fade, VideoStreamPlayer, and optional TTS dialogue from summer_generate_audio. - Use Case: Create a three-shot opening cinematic where a witch warns the player — one reference image drives all three 5-second kling clips, TTS dialogue plays over shot two, and the finished scene fades in and out at res://cinematics/Intro.tscn. ## Quick Start Ask your agent to generate the opening cutscene for the game using the cinematic-cutscene skill, starting with a reference image of the main character and chaining three 5-second shots.
