Video Generation

mold supports generating video clips using the LTX Video, LTX-2, and Wan 2.1/2.2 model families. Every LTX-2 and Wan checkpoint can chain multiple clips together for longer videos with scene-by-scene direction. A dev checkpoint renders its clips through the two-stage pipeline, so expect roughly twice the wall time per clip as a distilled one — stage 1 runs classifier-free guidance as two sequential forward passes.

Wan sequences

What crosses a wan clip boundary depends on the checkpoint, not the family: wan has no latent motion tail, so its smooth handoff is last-frame image conditioning. wan22-ti2v-5b and wan22-i2v-a14b continue across the seam; a text-to-video checkpoint concatenates independent clips (Join / Cut / Crossfade). Clip lengths sit on wan's 4k+1 grid, and --frames past the per-clip envelope auto-chains instead of failing.

mold run wan22-ti2v-5b:q8 "a paper boat drifting down a rain gutter" \
  --frames 100 --clip-frames 49

Single-clip generation

mold run ltx-2-19b-distilled:fp8 "a cat walks through autumn leaves" --frames 97

Wan Video

Wan renders single clips and multi-clip sequences, and defaults to MP4:

# 480p16 text-to-video (defaults: 81 frames @ 16 fps)
mold run wan21-t2v-1.3b "a red fox trotting through fresh snow, golden hour"

# Wan 2.2 A14B, 4-step Lightning tier (defaults: 81 frames @ 16 fps)
mold run wan22-t2v-a14b:q5 "a paper boat drifting down a rain gutter"

# A14B image-to-video from a still
mold run wan22-i2v-a14b:q5 "the balloon lifts off" --image balloon.png

Wan's frame grid is 4n+1 (49, 53, 81, 121, ...) from its VAE's 4x temporal compression, and dimensions must be multiples of 16 — except wan22-ti2v-5b, whose 2.2 VAE requires multiples of 32. A14B is a two-expert mixture: mold keeps one 14B expert resident at a time, so VRAM is the larger expert, not the sum. The :q5/:q4 tiers default to the checkpoint's trained 81 frames — automatic partial block offload fits them on a 24 GB card — while :q8 defaults to 73 frames and :fp8 to 45, their measured 24 GB envelopes. Wan checkpoints were tuned against a specific negative prompt that mold applies automatically when a request carries no negative at all; every surface shows it, editing replaces it, and clearing it (--no-negative on the CLI) sends a real empty negative. See Wan Video for variants, defaults, and limits.

Resolution and spatial tiling

LTX-2 normalizes RoPE pixel positions by a 2048px span, so a longer edge is outside the trained range even when the frame's area is small. A checkpoint that ships the spatial upsampler reaches past it by composing: stage 1 renders at half the requested size, the learned upsampler doubles it, and stage 2 refines the result over tiles each brought back inside the span. That puts the generation ceiling at 4096px on the long edge — exactly where a single halving stops landing stage 1 inside the span — and gives an output ladder of 1280x704, 1920x1088, 2560x1408 and 3840x2112. A checkpoint that renders in one pass stays capped at 2048px and 1920x1088.

The ladder stops at 3840x2112 because of the encoder, not the model: the bundled OpenH264 encoder refuses anything past 3840x2160, so 3840x2176 (what rounding 2160 up onto the /64 grid would give) generates correctly and then fails at save time.

There is nothing to assemble by hand: choose the output size and mold runs the composition. /api/models advertises the per-model max_pixels, max_axis_pixels and recommended_dimensions, so a picker never offers a rung the selected checkpoint would reject.

4K needs --spatial-tile 768 on a 24 GB card. Measured on an RTX 4090 with ltx-2-19b-distilled:fp8 at 25 frames: 1080p peaks at 18.4 GB, 1440p at 18.1 GB, and 4K completes at 18.2 GB with the smaller tile. With the default 1280px tile, 4K reaches the VAE decode and fails there with an out-of-memory error naming the phase — it does not silently render something smaller. Those are single-configuration measurements at 25 frames, not a support matrix; see LTX-2 → Resolution for the full numbers and caveats.

--spatial-tile (or MOLD_LTX2_SPATIAL_TILE, which mold serve reads) takes auto — the default, which tiles only where it buys something — off, or an explicit <px> / <px>:<overlap> in multiples of 32. auto engages exactly at the span and no earlier, so every resolution up to 1080p renders as it always has; off past the span is refused rather than quietly degraded. See LTX-2 → Spatial tiling.

Multi-prompt scripts (v2)

Direct any-length video scene-by-scene with a TOML script. Each prompt becomes a stage; each boundary has a transition (smooth, cut, or fade).

Canonical form

mold run --script shot.toml
mold run --script shot.toml --dry-run   # print stage summary, don't submit
mold chain validate shot.toml            # parse without submitting

Sugar form (uniform smooth chains)

mold run ltx-2-19b-distilled:fp8 \
  --prompt "a cat walks into the autumn forest" \
  --prompt "the forest opens to a clearing" \
  --prompt "a spaceship lands" \
  --frames-per-clip 97

Per-stage transitions or per-stage frames require --script.

Transitions

Mode	Behavior
smooth (default)	Motion-tail carryover — prompt change produces a visual morph between scenes.
cut	Fresh latent, no carryover. If the stage has source_image, it's used as an image-to-video seed.
fade	Cut + post-stitch alpha blend of fade_frames (default 8) on each side of the boundary.

Example shot.toml

schema = "mold.chain.v1"

[chain]
model = "ltx-2-19b-distilled:fp8"
width = 1216
height = 704
fps = 24
seed = 42
steps = 8
guidance = 3.0
strength = 1.0
motion_tail_frames = 25
output_format = "mp4"

[[stage]]
prompt = "a cat walks into the autumn forest"
frames = 97

[[stage]]
prompt = "the forest opens to a clearing"
frames = 49

[[stage]]
prompt = "a spaceship lands"
frames = 97
transition = "cut"

[[stage]]
prompt = "the cat looks up in wonder"
frames = 97
transition = "fade"
fade_frames = 12

Sequences in the apps

On web, desktop, and iPhone, multi-clip video is a setting rather than a separate page. In Create, set Output (beside Model) to Sequence and the composer becomes a clip rail: clip pills carrying a prompt and frame count, joined by seam pills. Seams are named in words — Smooth, Cut, and Fade 8f — and a click opens the seam editor with its fade-length stepper. LTX-Video has no motion tail, so its seams read Join. New clips default to the selected model's advertised frame count. Frame choices and timeline metadata include their duration at the live model FPS, such as 97f · 4.0s.

Sequences run as durable jobs in the same activity strip as ordinary prints. Desktop and web show each scene as a 16:9 filmstrip tile with its poster, render progress, and cache state. As soon as a stage finishes, its play control opens the raw scene in the main Create canvas while later stages continue rendering; Return to live render switches the canvas back to progress. Desktop and web can edit a finished sequence in place: its clips reload onto the rail, each pill shows cached (✓) versus re-render (↻), and Update sequence re-renders only from the earliest changed clip — transition and fade-length edits re-stitch with no re-render at all. Every settled sequence job is listed in Library ▸ History ▸ Sequences, and a sequence print in the Library uses Edit sequence as its primary desktop/web action (re-enter the original job with its cached clips). Duplicate as new starts a fresh sequence from the recorded clips. Desktop and web keep TOML import/export for mold.chain.v1 scripts under the composer's file tools. Web, desktop, and iPhone also offer Validate plan: it asks the currently selected authenticated host to normalize the live draft, then shows each clip's input/output frames, transition, conditioning inputs, warnings, and VRAM estimate when available. Validation creates no job and starts no download or inference work.

TUI chain composer

Press c from Create's navigation mode in mold tui to author a mold.chain.v1 script with per-stage prompts, frame counts, source images, and smooth / cut / fade transitions. See the TUI guide.

Capabilities endpoint

GET /api/capabilities/chain-limits?model=<name>

Returns per-model caps used by every sequence UI:

{
  "model": "ltx-2-19b-distilled:fp8",
  "frames_per_clip_cap": 97,
  "frames_per_clip_recommended": 97,
  "max_stages": 16,
  "max_total_frames": 1552,
  "fade_frames_max": 32,
  "transition_modes": ["smooth", "cut", "fade"],
  "quantization_family": "ltx2",
  "supports_audio": true,
  "supports_sequence": true
}

frames_per_clip_cap is the model's own clip size — what one generation renders when a long one-shot request is chained automatically (97 for LTX-2; for Wan the checkpoint's own manifest default over a 53-frame A14B / 121-frame floor, e.g. 121 for TI2V-5B) — so a sequence clip can never be longer than the clips the Duration slider would have produced. frames_per_clip_recommended follows the model's own default frame count — 97 for LTX-2, 25 for LTX-Video — so clients do not have to hardcode one. supports_sequence is model-specific and is also advertised per model on GET /api/models, so a picker never has to infer it from the checkpoint name. A family with no chain path reports false with a sequence_unsupported_reason. GET /api/models carries the matching per-model default_frames, default_fps, max_frames, and frame_step fields.