hook_variants

io.github.hermoso-ai/hermoso · Verify this server

HOOK MULTIPLIER: give ONE finished video ad N NEW OPENING HOOKS and get N complete edited versions to A/B test. WHAT CHANGES: a new opening shot over roughly the first 1.5-4 seconds (the hook ends at the source's first shot cut in that range, else at 3s; hookSeconds overrides). WHAT STAYS: everything after that point is the original footage, and the ENTIRE original soundtrack (voiceover, music, sound) plays under every version unchanged, so every version is the SAME length, aspect ratio and resolution as the source. Because the audio is kept, each hook is a VISUAL hook built to play under the words the source already says there: nobody in it talks to camera, it carries no on-screen text, and it does NOT write a new spoken hook line. Every version uses a DIFFERENT named hook mechanic chosen for the product (open mid-problem, before/after snap, object into frame, satisfying macro, pattern interrupt, POV, whip/snap-zoom, unexpected place, countdown to reveal); list_hooks describes them. Pass the video's file URL (a previous render, a job result, list_library, or upload_file for a local file). 1-5 versions, default 3. REFUSED FOR FREE, before anything is billed: a source over 120 seconds (trim it with post_edit first), one too short to leave 2 seconds of the original after a 1.5 second hook, an unreadable file, or a link to a social post rather than a video file (use clone_video to remake someone else's ad). COST: a small planning read, then each version is billed like fix_beat for the hook's seconds; the reply quotes credits per version, and dryRun:true returns the plan and the quote without rendering (pass that `plan` back to render exactly those hooks without planning again). Returns ONE JOB PER VERSION; call get_job on each until it reports done, and never describe a version before its URL arrives. Uses the workspace brand's product photo as a reference in hooks that show the product (productImage overrides; useBrand:false sends none).