MiniMax-H3 Multishot — Seamless Chain: multi-shot scenes that render as one continuous take (picture + audio)

v1.2 adds keyframes anywhere in the clip, not just the first and last frame — plus a hard mode for reference-to-video, and two controls ComfyUI reads but no stock node ever sets.Keyframes at any positionStock ComfyUI pins H3 keyframes to frame 0 and frame N−1, and raises <code>only first/last keyframe anchors are supported</code> for anything else. That turns out to be a positional-maths limit, not a model limit. Both stock cases are the same expression, because <code>sum(_video_t_spans(latent_t)) == FRAME_RESCALE * frame_count</code>:<pre><code>cond_t = text_len + FRAME_RESCALE * pixel_index</code></pre>which is defined for every frame. The new H3 Keyframes (any position) node takes up to six anchors, positioned as fractions (<code>0, 0.5, 1</code>) or absolute frame indices.Measured on an RTX 5090, 243 frames, one anchor at pixel frame 121: the rendered frame most resembling the anchor image was frame 122 — the requested position, off by one — reached by continuous motion with no cut (peak frame-to-frame change 2.3× the median), and the audio ran unbroken straight through it.The patch is applied in memory. It does not edit any ComfyUI file. It self-tests against the stock formula before committing and rolls itself back if first/last positions do not reproduce exactly — so a future ComfyUI change degrades to “interior anchors unavailable” rather than to broken renders.It moves, or it cuts — and your images decide whichAnchor images with a plausible camera path between them (same place, different angle or framing) make H3 interpolate: a real move that arrives on time. Images with no possible path — a kitchen and a diner — make it cut, then hold.That is not a limitation of this node. Stock first/last does exactly the same thing with such a pair; it is the model doing the only sensible thing. And the cut case has its own use: a timed shot change inside a single generation, which means the audio runs continuously across it, with no per-shot voice re-derivation and no seam.Reference-to-video groundworkThe two nodes below are what reference workflows need. The hard-mode graph itself is built but has not been rendered yet, so it is not in this release — it ships when it has a render behind it.References and keyframes are mutually exclusive, in ComfyUI core rather than by choice here. <code>model_base.py</code> assigns <code>cond_video_latents</code> for refs, discarding any keyframe latents while the keyframe layout rows survive — the packed sequence then desyncs into a shape-mismatch crash. There is no “reference images plus start frame” mode. Pick one per shot.Two things that are easy to get wrong and are documented on the canvas: reference tags are 1-based in the prompt (<code>&lt;Picture 1&gt;</code>) while the input slots are 0-based (<code>ref_image_0</code>); and a reference video with a soundtrack consumes an <code>&lt;Audio j&gt;</code> ordinal before your standalone clips.Two more nodes<ul><li>H3 Reference Audio (stereo guard) — the audio VAE encodes <code>[B, 2, L]</code> and the layout reserves exactly two channels, so a mono reference produces half the rows it reserved and dies deep inside the model with no useful error. Nothing in stock converts it. This forces stereo 32 kHz and trims length.</li><li>H3 Condition Strength — exposes <code>minimax_visual_cond_noise_aug</code> and <code>minimax_audio_cond_noise_aug</code>, which core reads but no stock node writes, so they sat at 0.999 / 1.0 permanently. They blend noise into the conditioning latents and set the timestep the condition rows occupy: an anchor-strength dial.</li></ul>Still in the box from v1.1<ul><li>Image-to-video via an optional <code>start_image</code>.</li><li>~4× faster on 32 GB cards. The text encoder is evicted before sampling — encoder (~16.5 GB) and DiT (~25 GB) do not co-fit, so the DiT was loading partially and streaming ~19 GB from system RAM every step. RTX 5090: ~60 min to ~15 min, same render. The keyframes node does this too.</li><li>Long-form memory sampler for 2–5 minute pieces.</li><li>Audio seams crossfaded; <code>seed_per_shot</code> on by default.</li></ul>Links<ul><li><a target="_blank" rel="ugc" href="https://github.com/jlucasmcrell/ComfyUI-H3-Multishot">Node pack on GitHub</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/joeygambino/MiniMax-H3-GGUF">H3 DiT GGUF</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/joeygambino/MiniMax-H3-encoder-GGUF">Text encoder GGUF (+ mmproj)</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3">VAEs</a></li></ul>SupportEverything I publish is free and stays free. If it saved you a night of debugging, tips keep the 5090 warm: <a target="_blank" rel="ugc" href="https://ko-fi.com/joeygambino">Ko-fi</a> · <a target="_blank" rel="ugc" href="https://github.com/sponsors/jlucasmcrell">GitHub Sponsors</a> · <a target="_blank" rel="ugc" href="https://liberapay.com/joeygambino">Liberapay</a>.

minimaxH3MultishotSeamlessChain_v12.zip

v1.2

v2.6.3

v2.6 Extend Take

v.2.2.5

v2.2.4 Multi-Ref+Vid-Ref

v2.2.3 More Mem Fix

v2.2.2 - Glasses?

v2.2.1 - Bugfix

v2.2 - Big Update

v2.1.2

v2.1.1

v2.1 Bugs in my bugfixes

v2.0 - Seamless Chain

v1.5 Auto-Voice Ref+LoRAs

v1.4.1 - Bugfix release

v1.4

v1.3

v1.2.1

v1.1

v1.0

Files with SHA256: 6e972e7aaaf7bfe2a7b2426c56d4bad2ee128e1249c3eada44ee88d302d57a23