[Legacy] H3 Continuum Advanced

H3 Continuum Advanced is the settings bundle that paired with the compact H3 Continuum Sampler. Same idea as its V3 sibling: instead of plastering every continuation toggle across the sampler, the pack routes them through one node whose single advanced output plugs into the sampler's advanced input. It's legacy now - the modern V3.3 sampler absorbed these settings as native widgets - but you'll still meet it in V1/V2-era workflow files.

The settings it owns are the continuation knobs you touch rarely but need when you do:

• audio_continuity (true) - carry the previous raw audio latent into continuation chunks.
• exact_total_duration (true) - trim or minimally pad the assembled result to exactly chunks × seconds × 24fps.
• reroll_from_chunk (0) - 0 reuses every matching accepted chunk from the optional session; N preserves chunks before N and regenerates N onward.
• reroll_nonce (0) - change only when regenerating an explicit chunk and you want a new variation with identical settings.
• strict_compatibility (true) - proactive native-H3 API checks; marked payload failures always stop.
• diagnostics (Basic) - report verbosity only; never changes generation or seam correction.
• debug (false), show_preview (true).

Optional inputs let you thread continuation state through: last_frame, session, initial_state, and prompt_plan (from the Prompt Plan Preview node).

Install

cd ComfyUI/custom_nodes
git clone https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Continuum.git

or via ComfyUI Manager ("H3 Continuum"), then restart. No extra pip packages; needs a MiniMax H3 ComfyUI environment (weights not bundled) and ComfyUI v0.32.0+.

Practical take: most of these defaults are the right call and you can leave the node untouched. The two you might actually flip are reroll_from_chunk (fix just the ending without regenerating the whole run) and audio_continuity = false (when you want to isolate or fully replace the generated audio rather than inherit it). If you're building a new workflow, skip this node entirely and use H3 Continuum Sampler V3.3, where audio_continuity and reroll controls sit directly on the sampler and Run Storage handles resume automatically.