Luma Ray 3.2: what's new, price and how to use it
Luma released Ray 3.2 on 9 June 2026: 16 keyframes, native 16-bit HDR, 5 or 10 s clips to 1080p, no audio. On Versely since 18 October, from 12 credits.
Every guide, comparison and workflow we’ve published on AI Video Generator.
213 articles — page 1 of 9
Luma released Ray 3.2 on 9 June 2026: 16 keyframes, native 16-bit HDR, 5 or 10 s clips to 1080p, no audio. On Versely since 18 October, from 12 credits.
Grok Imagine Video 1.5: xAI API in June 2026, native 1080p from 31 July, 1-15 s, lip-synced audio, from $0.08/s. Lite $0.02/s. On Versely from 7 credits.
Generate a Seedance 2.5 clip in Versely on the web and in the iOS app: text, image or reference to video, 4 to 30 seconds, native audio, 720p. Now 50% off.
A paced slideshow of generated stills beats generated video on revision cost and variants. It loses on anything that must be demonstrated. One rule decides.
A short dissolve at the join is the standard fix for exposure and identity drift between chained clips. Dissolve lengths, and when a hard cut still wins.
Which camera moves survive as text, which get silently approximated, and what a motion-reference clip costs you in framing freedom. With a three-bucket rule.
Three ways to carry a shot past a model's duration ceiling, compared on look-hold, camera continuity, credit cost, and what a bad segment costs to fix.
An elimination protocol that takes 296 models to 8 in five ordered gates: input modality, duration ceiling, audio, licence, credit band. Refill quarterly.
Black Forest Labs trained FLUX 3 jointly on image, video and audio with 20s synced sound. Which Flux still-image workflows should absorb it, and which won't.
Frame interpolation invents frames between real ones. Native frame rate computes each moment. A 60 fps label can be either. Contact and occlusion break it.
A match cut hides the join between two generations by carrying a shape or a motion across it. How to plan the pair in the prompt and which frames to trim.
Directing three shots inside one generation versus generating three clips and cutting them, compared on continuity, per-shot control and what a retake costs.
Ramping buys pacing without extra generations, but only inside the frame budget you actually have. Safe ratios, where interpolation helps, where it smears.
Going straight to 4K amplifies artifacts you could have removed cheaply. The pass order that works, and what each pass can and cannot fix.
Shimmer is usually a frame-rate problem: too many frames starve temporal context. Rates that hold, where the deflicker pass goes, and how to spot compression.
Some camera directives are real parameters. Others are prose the model approximates and quietly drops. How to tell which is which, and test for it.
'It doesn't look consistent' isn't one problem. It's four — identity, wardrobe, set and lighting — each with a different cause and a different fix.
Continuation prompting explained: hand a model the end of an existing clip plus an instruction for what happens next, and extension becomes direction.
Why whip pans and crash zooms break text-to-video prompts, what Wan 2.2 docs reveal about how camera motion is learned, and when to switch to motion transfer.
Restarting a stalled multi-scene render throws away every completed scene. The real state machine, the stuck threshold, and what resume actually resets.
Why image-to-video generations fail before the prompt even runs: reference resolution floors, framing, motion blur, and the clean-cutout trap explained.
Grok Imagine Video 1.5 references: 7 at the 31 July 2026 launch, 14 in xAI's docs now, up to 720p, preset voices only. Versely's reference row takes 7.
Every video model caps clip length. The technique that holds a look across a longer scene isn't a longer prompt — it's defining frames and chaining the joins.
Kling 3 Turbo T2V and I2V on Versely are audio off, 22 credits min, 8 credits/s discounted. Plan silent video plus captions, not bundled dialogue.