Ten LTX-2.5 clips generated 2026-08-17 on one B200, against the H3
clips already on disk. Same 768×1344 at 24fps, same in-model audio, same concepts
and verbatim dialogue. Clips are full quality — nothing re-encoded — so skin
texture is judgeable. Unmute to check the audio; both models
generate it in-model.
~$0.13LTX per 15s clip
~$0.97H3 per 15s clip
10 / 10LTX clips succeeded
66smean LTX gen time
What to look for. The result splits by conditioning path, and
that matters more than the price.
Image-conditioned (green tag): LTX is genuinely competitive —
identity, set, wardrobe and props hold for the whole clip. H3 still runs cooler
and more matte, which is the look the house prompts were tuned for, and it holds
framing where LTX wanders.
Text-only (red tag): LTX drifts. In walk_p1 the
person changes and the lighting jumps mid-shot. H3's ref2va holds a
presenter across scene changes and LTX-distilled has no equivalent —
that gap, not the cost, is what decides adoption for recurring-cast work.
grocery_trailer_voice
image-conditioned
Best case for LTX. Identity, kitchen, wardrobe and props hold for the whole clip, and the comedy arc lands: amused → wide-eyed → the mock-trailer bellow → collapsing into laughter.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 243f · 10.12s · 32000Hz · 1.8MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 241f · 10.04s · 48000Hz · 3.4MB
part1_hook
image-conditioned
15s image-conditioned. Identity holds across all 361 frames. LTX pulls in closer and is more animated; H3 holds the reference framing more strictly.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 362f · 15.08s · 32000Hz · 2.1MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 361f · 15.04s · 48000Hz · 5.0MB
smoke_short
image-conditioned
Short control clip. Note LTX drifts off-prompt near the end — stops addressing camera and drops the earbud the prompt named.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 107f · 4.46s · 32000Hz · 0.6MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 105f · 4.38s · 48000Hz · 2.1MB
walk_p1
text-only
THE FAILURE CASE. Watch the LTX side: the person visibly changes partway through, and the lighting jumps from cool daylight to hard golden sunset mid-shot. H3's ref2va holds one presenter and one lighting state for 15s.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 362f · 15.08s · 32000Hz · 4.2MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 361f · 15.04s · 48000Hz · 15.8MB
part2_reveal
text-only
Text-only bedroom. H3 uses ref2va to keep the same presenter in a new room; LTX has no equivalent, so this is a fresh person each time.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 362f · 15.08s · 32000Hz · 2.1MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 361f · 15.04s · 48000Hz · 4.2MB
part3_payoff
text-only
Text-only bathroom. Same structural gap as above.
MiniMax-H34×B200 · 24 steps · ~$0.97
768×1344 · 362f · 15.08s · 32000Hz · 2.3MB
LTX-2.51×B200 · 8+4 steps · ~$0.13
768×1344 · 361f · 15.04s · 48000Hz · 9.5MB
bedroom_morning_show
text-only
Incidental win for LTX: it came out correctly PORTRAIT. H3's only version of this concept is 1344×768 LANDSCAPE — the ref2va + aspect_ratio:auto trap, a full-price unusable clip.