Other models
Luma Ray3.2, Grok Imagine 1.5, LTX-2.5, PixVerse V6 and C1, Midjourney video, and what the Sora 2 shutdown means for you.
These models are useful for specific jobs, not as your main generator. Pick Grok Imagine for cheap short loops, LTX-2.5 or Luma Ray3.2 for HDR and finishing work, PixVerse for cheap clips with audio, and Midjourney only if you already pay for it for images. Sora 2 is gone.
Quick pick
Prices checked on 15 Oct 2026.
| Model | Use it for | Max length | Audio | Cheapest way in |
|---|---|---|---|---|
| Luma Ray3.2 | HDR footage, many keyframes, reframing | 20 s | Not mentioned (treat as none) | Plus: 30 USD/month |
| Grok Imagine 1.5 | Short loops, simple animation, quick tests | 15 s | Yes, with lip sync | Included in SuperGrok (30 USD/month) |
| LTX-2.5 | Open weights, upscaling, alpha mattes, HDR | about 20 s (audio to video) | Yes | Open weights, or LTX Studio Free (800 credits once) |
| PixVerse V6, C1 | Cheap clips with audio, storyboard to video (C1) | 15 s | Yes | API: about 0.20 USD for 5 s at 720p without audio |
| Midjourney V1 | Animating your Midjourney images | 21 s with extends | No | Basic: 10 USD/month (third party source) |
| Sora 2 | Nothing, it has shut down |
Luma Ray3.2
At a glance
| Item | Detail |
|---|---|
| Maker | Luma AI |
| Current versions | Ray3.2 (released 9 Jun 2026), Ray3.14 |
| Max duration | Up to 20 s per clip |
| Resolution | Up to 1080p, native HDR with 16 bit EXR |
| Audio | Not mentioned in the launch post. Treat it as silent. |
| References | Up to 16 keyframes per clip. Reframe (change aspect ratio or background). Facial performance tracking for up to 8 faces. |
| App and API | lumalabs.ai (Dream Machine) and the Luma API |
| Price | Plus 30 USD/month (25 USD annual), 10,000 credits. A 5 s Ray3.2 clip: 100 credits at 720p, 400 at 1080p. On the API: 0.30 USD (720p) or 1.20 USD (1080p) for 5 s. |
Best for and weak at
- Best for: HDR and color graded pipelines, shots that need many keyframes, and reframing existing clips for other aspect ratios (official features).
- Weak at: sound, and community attention. We found almost no creator posts. One low reach post rated Ray3.14 about as good as Seedance 1.5 Pro, without sound.
- Beware: the X account @LumaAIX is unrelated spam.
How to prompt it
Luma has no official prompt guide that we could read. Use the general structure: shot and camera, subject, action, setting, light. Use keyframes to pin the moments that matter instead of describing them in text.
Slow aerial descent over a terraced rice field in the mountains at sunrise, mist in the valleys,
a single farmer walking along a ridge. Soft golden light, deep green and gold palette.Settings that save credits
- Draft mode costs 20 credits for 5 s, against 400 at 1080p.
- HDR doubles the cost, HDR plus EXR triples it. Turn it on for the final only.
- Free tier, watermark and rollover terms were not on Luma's pricing page. Third party sources say monthly credits do not roll over.
Grok Imagine 1.5
At a glance
| Item | Detail |
|---|---|
| Maker | xAI |
| Current versions | Grok Imagine Video 1.5 and 1.5 Lite (generally available from about mid June 2026). The older model is used for editing. |
| Max duration | 1 to 15 s (default 8 s) |
| Resolution | 480p (default), 720p, 1080p. 1.5 renders 1080p natively for text and image to video. Reference to video is up to 720p. Lite upscales 720p to 1080p. |
| Audio | On by default, with lip synced speech. Up to 3 speakers with preset voices. |
| References | Up to 14 images and 3 voice references. First, last and middle frames can be pinned. |
| App and API | Grok app and X (SuperGrok 30 USD/month includes video; bundled with X Premium per users). xAI API. |
| Price per second (API) | 1.5: 0.08 USD (480p), 0.14 (720p), 0.25 (1080p). 1.5 Lite: 0.02, 0.03, 0.14. |
An 8 s clip at 720p costs about 1.12 USD on 1.5 and 0.24 USD on Lite. Vietnam availability is not confirmed.
Best for and weak at
- Best for: quick, cheap tests and simple loops. One organic creator animates a mascot at 480p, 1:1, without sound. Many hobbyists start here because it comes with X Premium.
- Weak at: motion. Organic testers call movement "floaty" and slow, and in side by side tests it lost to Seedance. Still frames look better than the movement. One user reported that prompts with references became hidden.
How to prompt it
xAI has no dedicated video prompt guide. Its examples describe motion and camera explicitly. Text to video first makes a still frame, then animates it, so describe the opening image clearly.
A round orange robot mascot on a white background waves its left arm, blinks twice and bounces once.
Static camera, flat 2D cartoon style, seamless loop. No sound.Settings that save credits
- Use 1.5 Lite at 480p or 720p for tests (0.02 to 0.03 USD/second).
- Turn audio off (
generate_audio=False) for loops that do not need sound. - Image inputs cost 0.01 USD each on the API.
LTX-2.5
At a glance
| Item | Detail |
|---|---|
| Maker | Lightricks |
| Current versions | LTX-2.5 (release date not on its page), LTX-2.3 (about March 2026) |
| Open weights | Yes. Free commercial use under 10M USD revenue and a 16 GB VRAM minimum, per community posts. |
| Features | Native multi shot, automatic duration, native HDR (16 bit EXR in and out), alpha mattes (AlphaGen) |
| Max duration | About 20 s per request for audio to video |
| Resolution | Up to 4K |
| Audio | Joint audio generation and audio to video |
| App and API | LTX Studio and the LTX API. Also on Runway. |
| Price per second (API) | LTX-2.5 Fast: 0.09 USD (720p), 0.13 (1080p), 0.30 (4K). LTX-2.5 Pro: 0.12, 0.17, 0.39. LTX-2.3 Fast from 0.03 USD at 720p. |
LTX Studio plans: Free with 800 credits once (personal use only), Lite 15 USD/month (no watermark), Standard 35 USD (commercial license, adds Kling 3.0 Pro and Seedance 2.0), Pro 125 USD (adds Veo 3.1).
Best for and weak at
- Best for: utility work. Creators use it for upscaling drafts from other models, alpha matting, converting 8 bit to 16 bit EXR, and cheap multi person lip sync (organic reports).
- Weak at: hero shots with action. Testers do not use it as their main model for action scenes.
How to prompt it
LTX's official guide asks for one flowing paragraph in the present tense, 4 to 8 sentences: shot framing, scene, action, character, camera movement and audio, with dialogue in quotes (from a search snippet, the page blocked our reader).
A close up of an old clockmaker at his bench, lit by a single desk lamp. He lifts a tiny gear with tweezers
and sets it into an open pocket watch. The camera slowly pushes in toward his hands. He whispers, "There you are."
The room is quiet except for the ticking of a dozen clocks.Settings that save credits
- Draft on LTX-2.3 Fast (0.03 USD/second at 720p), then use 2.5 for the final.
- Run the open weights if you have a 16 GB GPU.
PixVerse V6 and C1
At a glance
| Item | Detail |
|---|---|
| Maker | PixVerse |
| Current versions | V6 (30 Mar 2026), the general model. C1 (7 Apr 2026), for film production. R1 and R2 are real time world models, not clip generators. |
| Max duration | 1 to 15 s |
| Resolution | 360p, 540p, 720p, 1080p. Ratios from 21:9 to 9:16. |
| Audio | Optional native audio |
| References | Fusion references, including video references. C1 also turns a multi panel storyboard into video. First and last frame on both. |
| Other features | V6: multi clip and extend. C1: no extend. Lip sync, sound effects, restyle and motion copy are separate API features. |
| App and API | pixverse.ai (also hosts Seedance 2.5 and MiniMax H3) and the PixVerse API |
| Price | API: V6 at 720p is 9 credits per second without audio, 12 with audio. PixVerse's own example: 1 USD buys 5 videos (V6, 720p, 5 s, no audio). |
Consumer plan prices did not load. Third party sites list Standard at 10 USD/month for 1,200 credits. Not confirmed.
Best for and weak at
- Best for: cheap clips with audio (about 0.27 USD for 5 s at 720p with audio on the API), and storyboard to video on C1.
- Weak at: PixVerse itself says complex action, multilingual dialogue and product accuracy may need retries. Community attention in 2026 is on R2, not on V6 or C1.
How to prompt it
PixVerse's official tips: use literal physical descriptions, not metaphors. Order: subject, action, camera path, lighting, material, scene changes, dialogue, sound effects. Write dialogue as speaker, tone in parentheses, then the line. For multi clip continuity, repeat the same character, product, lighting and location words in every shot.
A young woman in a white linen shirt pours tea from a clay pot into a small cup on a wooden table.
Camera moves slowly from her hands up to her face. Soft morning light from a window on the left, steam rising.
Woman (calm, quiet): "This one is from my grandmother's farm."
Sound: tea pouring, a cup set down on wood.Settings that save credits
- Draft at low resolution without audio, then make the final at 1080p with audio (PixVerse's own advice).
- Change one thing per retry.
- Video references double the credit cost on V6.
Midjourney video V1
At a glance
| Item | Detail |
|---|---|
| Maker | Midjourney |
| Current versions | V1 (June 2025), image to video only. No confirmed 2026 video model. |
| Max duration | 5 s base. Extend adds 4 s up to 4 times, 21 s in total. |
| Resolution | SD, and HD on Standard, Pro and Mega plans in Fast mode |
| Audio | None |
| App and API | Midjourney web app. No API. |
| Price | Basic 10 USD, Standard 30, Pro 60, Mega 120 per month (third party source). One HD clip uses about 7 GPU minutes, about 0.47 USD of extra fast time. Relax mode covers SD video on Pro and Mega. |
Best for and weak at
- Best for: animating images you already made in Midjourney. Pros still use Midjourney for look development and storyboards, then animate in other models.
- Weak at: sound (none), length, and text to video (not supported). A "V8 Video" model is claimed only on a competitor's site and is not confirmed.
How to prompt it
Start from an image, then add a short motion prompt and parameters: --motion low or --motion high, --raw, --loop, --end for an end frame, --bs 1, 2 or 4 for batch size.
the paper boat drifts slowly down the rain gutter, raindrops hitting the water around it --motion low --bs 1Settings that save credits
- Use
--bs 1. A batch of 4 costs 4 times the GPU time. - Every extend costs as much as a new generation.
- Use Relax mode for SD video on Pro or Mega.
Sora 2 has shut down
- The Sora app and website closed on 26 Apr 2026.
- The Sora 2 API and the Videos API closed on 24 Sep 2026.
- OpenAI says there is no one to one replacement API. Leftover ChatGPT or Sora credits can be used on Codex.
- Data is deleted after the final export window.
Do not pay for Sora 2 anywhere
Some aggregators still list "Sora 2" on their pages. Those listings are stale at best. Do not buy a plan for Sora access.
OpenAI's Sora 2 prompting guide is still worth reading for general craft: write prose, camera and action beats separately, give each shot one clear camera move and one clear subject action, and keep clips short. The ideas apply to every model on this site.
Sources
- Luma Ray3.2 announcement
- Luma pricing
- Luma API pricing
- xAI video generation docs
- xAI video guide
- xAI pricing
- xAI consumer pricing
- LTX-2.5
- LTX API pricing
- LTX Studio pricing
- LTX prompting guide
- PixVerse V6 API
- PixVerse C1 API
- PixVerse API pricing
- PixVerse V6 prompt tips
- Midjourney video docs
- OpenAI: Sora discontinuation
- OpenAI video generation guide
- Sora 2 prompting guide