H3 workflow trick: use Add Guide nodes to insert visual references without typing
ForsakenAd1228 · reddit · 2026-09-04
A user shares a practical H3 video-generation workflow using the Add Guide node to avoid typing long references:
- Basic use: insert a wide shot of the room + characters on frame 0 via Add Guide, then start the prompt with "At 00:00.050, cut to ...". Less useful for I2V (which has a first-frame input), but in ref2v it is more reliable and far less typing than the official ref2va workflow.
- Double Add Guide for clip extension: to extend a clip that ended on a face close-up while showing the room again, add one guide with the full room/character on frame 0 and a second with the last 5 frames of the previous clip from frame 1. Prompting "At 00:00.050, cut to <what happens at the end of the first clip>" yields a seamless transition with zero typed references.
- Works with higher-quality I2V too; the trade-off is the reference image appearing on the first frame, which is fine once clips go through a video editor.
More from coding & agent
- Reading code in the AI era: one dev reads all day to steer coding agents, misses provider outage — JFPuget · 2026-09-04
- Why cloud-based AI agents fit long-running development tasks, per OpenRouter — thisiskp_ · 2026-09-04
- LlamaIndex launches Turbo extraction tier: 4x faster at 3.7s median per page — llama_index · 2026-09-04
- DuoSteer: steering LLM attention heads fixes code vulnerabilities without hurting correctness — ZiyuYao · 2026-09-04
- I still read code — but only to steer my coding agent better, engineer says — JFPuget · 2026-09-04
- Models Keep Improving, but the Engineering Around Them Hasn't Kept Up — Meher_Nolan · 2026-09-04