
LTX 2.5 Fast API
ltx-2.5-fastLTX 2.5 Fast is the speed-and-cost tier of Lightricks' LTX-2.5 video model: it turns a text prompt, a still image or a voice or music track into a 6–20 second video with matching sound at up to 1440p, through one API call and without your own GPUs. Call ltx-2.5-fast from $0.081/s at 720p, 10% below Lightricks' API list price; failed tasks are not charged.
Generate Video
Result
Generated video will appear here
Provide URLs and click Generate
Three Modes, One Model ID
A prompt alone is text-to-video; add image_url for image-to-video from a first frame; add audio_url for audio-to-video. The mode follows what you send
Audio-to-Video
Your MP3, WAV, M4A, AAC, OGG or FLAC track (2 s or longer) becomes the soundtrack and the picture is generated to match it. Length follows the audio, or set 2–20 s; billed to 0.01 s
720p, 1080p or 1440p
At 24, 25, 48 or 50 fps. Text-to-video takes 16:9, 9:16, 1:1, 3:2 or 2:3; image-to-video keeps the shape of your first frame
6–20 Seconds
Even seconds from 6 to 20 at 720p / 1080p with 24 or 25 fps; 6, 8 or 10 s at 1440p or 48 / 50 fps. Sound is generated with every clip
10% Below the ltx.io List Price
$0.081 / $0.117 / $0.171 per second at 720p / 1080p / 1440p, against $0.09 / $0.13 / $0.19 on Lightricks’ API pricing page. Frame rate does not change the price
Limits, Stated Plainly
No 4K, last frame, reference images or video input here. Content review or a busy upstream can fail a task; failed tasks are not charged
LTX 2.5 Fast API: text, image or audio to video with sound, from $0.081 per second
LTX 2.5 Fast is the speed-and-cost tier of LTX-2.5, Lightricks' open-weights video and audio model of 11 August 2026, and it turns a prompt, a still image or a voice or music track into a 6–20 second clip with sound at up to 1440p.
Use it when you want to animate a photo, make a talking-head or music-driven clip from an audio file, or generate short social videos from text without running a 22B model on your own GPUs.
On APIMODELS the LTX 2.5 Fast API is the model id ltx-2.5-fast on POST /v1/video/generations, from $0.081 per second at 720p, 10% below Lightricks’ list price; a failed task is not charged.
- Model ID
ltx-2.5-fast- Modes
- Text · image · audio to video
- Resolution
- 720p · 1080p · 1440p
- Length
- 6–20 s
- Price
- from $0.081/s
LTX 2.5 Fast examples: five real clips from our tests
Each clip is one LTX 2.5 Fast request at 720p, run once on 20 November 2026 and shown as it came back. Open a card for the exact prompt and any input image.
What each LTX 2.5 Fast clip cost and how long it took, queue included: the dialogue, the three-shot ride, the vertical product clip and the photo-to-video clip are 6 seconds each, $0.486 apiece, and came back in 41, 27, 64 and 73 seconds. The voice-track clip used a 4.8-second recording, was billed for 4.81 seconds ($0.39) and took 86 seconds.
What did not go to plan: in the dialogue the cook sits at the same table as the customer instead of standing behind the counter; in the product clip the ice looks like white lumps and the bottle changes shape between shots; in the photo clip the camera drifts instead of orbiting and the soundtrack is almost silent. The voice-track clip kept our recording unchanged, and both spoken lines in the dialogue came out word for word.
Last bowl of the night, on the house. / You say that every night.Prompt (exact text sent)
Static medium shot inside a tiny ramen shop at night, steam rising from a pot behind the counter, warm paper lanterns overhead. A cook in his fifties with a white headband wipes the counter, looks up at a young woman sitting across from him and says, "Last bowl of the night, on the house." She laughs and replies, "You say that every night." Sound: broth simmering, light rain on the window, their two voices.Prompt (exact text sent)
A cyclist in a yellow jacket rides along a coastal cliff road at sunrise. Side tracking shot keeping pace with the rider, then a cut to a low close-up of the front wheel splashing through a puddle, then a wide high shot of the road curving along the sea. Sound: wind rushing, the chain clicking, waves crashing below.Prompt (exact text sent)
Vertical close-up of a glass bottle of cold brew coffee on a wet black stone with cold mist around it. A hand reaches in, twists off the cap with a crisp hiss and pours the coffee over ice in a tall glass while the camera slowly pushes in. Sound: the cap hissing, ice cracking, coffee pouring.
Prompt (exact text sent)
Steam rises gently from the mug, warm morning sunlight slowly moves across the oak table, and the camera orbits slowly to the right around the mug. Sound: quiet cafe ambience and birds outside the window.Welcome back to the show. Today we are testing how well one single photo can talk.Reference images
Prompt (exact text sent)
A podcast host talks warmly to the camera in her home studio, natural head movements and small hand gestures that follow her speech, static medium shot, soft warm lamp light.How to use the LTX 2.5 Fast API in 4 steps
An LTX 2.5 Fast job is one request to create the task and a second to read the result; the mode follows the fields you send.
Create an API key
Sign up at apimodels.app and create a key in the console. The same key and balance work for every model on the site; no Lightricks account is needed.
Pick the inputs, which pick the mode
LTX 2.5 Fast reads the mode from your inputs: prompt only for text-to-video; add image_url for image-to-video from a first frame; add audio_url for audio-to-video. A prompt is required in all three.
Set resolution, fps and length
LTX 2.5 Fast takes resolution 720p, 1080p or 1440p; fps 24, 25, 48 or 50; duration 6–20 s in even steps (6, 8 or 10 at 1440p or 48 / 50 fps). For audio-to-video, leave duration out to follow the audio.
Poll for the video
Every LTX 2.5 Fast request is asynchronous: POST /v1/video/generations returns a taskId. Poll GET /v1/video/generations?task_id= until state is completed and download data.resultUrls[0], or pass callback_url. Files are kept 30 days.
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "ltx-2.5-fast",
"prompt": "Close-up of an old fisherman mending a net on a wooden pier at dawn, low warm sun on his face. He pulls the cord tight, looks up and says, \"Storm is coming.\" The camera slowly pushes in while gulls cry and waves slap the posts.",
"resolution": "1080p",
"fps": 24,
"duration": 10,
"aspect_ratio": "16:9"
}'LTX 2.5 Fast image-to-video, audio-to-video, Python and polling examples are on the LTX 2.5 API docs page.
What LTX 2.5 Fast makes: text to video, image to video and audio to video
Seven common LTX 2.5 Fast jobs, each with the fields to send and the limit that applies to it.
Short social clips from a script
LTX 2.5 Fast makes a social clip from one paragraph that covers the shot, the action, the camera and the sound; set aspect_ratio 9:16, 1:1 or 16:9. Each clip comes back with its own soundtrack; 10 seconds at 720p is $0.81.
Limit: 20 seconds per clip, and Lightricks says on-screen text is still not reliable, so add titles in editing.
promptaspect_ratioAnimate a photo from its first frame
LTX 2.5 Fast animates a photo when you put it in image_url and describe what moves. The clip opens on that exact picture and keeps its aspect ratio; 6 seconds at 1080p is $0.702.
Limit: first frame only; there is no last frame, so you cannot pin where the shot ends.
image_urlProduct shot from a packshot
For a product shot, LTX 2.5 Fast takes a clean packshot as the first frame; ask for a slow push-in, a turn or a hand picking it up, with ambient sound. 1440p is available for 6, 8 or 10 seconds.
Limit: printed labels and logos can drift as the camera moves; check every frame before you publish.
image_urlresolutionTalking head or lip sync from a voice track
LTX 2.5 Fast makes a talking head from a voice recording in audio_url and a portrait in image_url; the recording becomes the soundtrack and the picture is generated to match it. 12.5 seconds at 1080p is $1.4625.
Limit: up to 20 seconds per clip (10 at 1440p or 48 / 50 fps), and we have not published our own lip-sync tests yet.
audio_urlimage_urlA music-driven clip from a song
LTX 2.5 Fast builds a music-driven clip from a 2–20 second excerpt in audio_url and a scene description; set duration to cut a shorter piece from the start. Billing follows the output length to 0.01 s.
Limit: without an image the clip is always 16:9; send a first frame in the shape you want for vertical video.
audio_urldurationSeveral shots in one clip
Lightricks says LTX-2.5 can cut between connected shots in one generation; its guide asks you to name each cut in prose and keep to 2–4 shots, with 20 seconds as the longest clip here.
Limit: this is the vendor’s claim; we have not tested multi-shot prompts through this API yet.
promptdurationDrafts before a Pro render
LTX 2.5 Fast is where prompts get tried: 720p costs $0.486 per 6-second take, then send the chosen prompt to ltx-2.5-pro with the same fields.
Limit: Pro stops at 10 seconds and 1080p and has no audio-to-video here, so long or audio-driven shots stay on Fast.
ltx-2.5-proWho LTX 2.5 Fast is for, and who should use something else
The cases LTX 2.5 Fast fits, and the cases where another model or route is the better answer.
Developers adding video to an app
LTX 2.5 Fast gives you one key, one endpoint and a per-second price you can compute before the request, with audio included and no GPUs to run.
Creators with a voice-over or a song
LTX 2.5 Fast audio-to-video keeps your track and builds the picture around it, for talking heads, explainers and music clips up to 20 seconds.
Marketers making many short variants
LTX 2.5 Fast offers five aspect ratios in text-to-video, 1440p when the platform allows it, and a price per clip that scales with seconds only.
Teams that cannot run the open weights
The open model needs your own GPUs and pipeline; here LTX 2.5 Fast is one HTTP call, billed per second of output.
You need 4K, a last frame or a silent file
LTX 2.5 Fast on apimodels.app has none of these. Lightricks’ own LTX API and fal.ai offer 4K, an end frame and generate_audio false; otherwise strip the audio in editing.
You need clips longer than 20 seconds
LTX 2.5 Fast stops at 20 seconds. Seedance 2.5 and Wan 3.0 Video on this site go to 30 seconds in one clip.
You need several references to hold a character
LTX 2.5 Fast takes no reference images or reference videos. MiniMax H3 takes up to 9 images, 3 videos and 3 audio clips; Seedance 2.5 up to 30, 10 and 10.
You want to fine-tune or train a LoRA
The hosted LTX 2.5 Fast API cannot load your adapters. Self-host the open weights from Hugging Face, where Lightricks publishes the trainable checkpoint.
LTX 2.5 Fast vs LTX 2.5 Pro vs LTX 2.3 on apimodels.app
The three LTX models on this site, with the limits and prices each one actually accepts here.
| LTX 2.5 Fast | LTX 2.5 Pro | LTX 2.3 | |
|---|---|---|---|
| Model id | ltx-2.5-fast | ltx-2.5-pro | ltx-2.3 |
| Modes | Text, first-frame image, audio to video | Text, first-frame image to video | Text, one image to video |
| Resolution | 720p, 1080p, 1440p | 720p, 1080p | 480p, 720p, 1080p |
| Length | 6–20 s in even steps at 720p / 1080p and 24 / 25 fps; 6, 8 or 10 s at 1440p or 48 / 50 fps | 6, 8 or 10 s | Text 5–15 s, image 5–20 s, any whole second |
| Frame rate | 24, 25, 48, 50 fps | 24, 25, 48, 50 fps | Not selectable |
| Aspect ratio (text-to-video) | 16:9, 9:16, 1:1, 3:2, 2:3 | 16:9, 9:16, 1:1, 3:2, 2:3 | 16:9, 9:16 |
| Audio-to-video | Yes, 2–20 s, billed to 0.01 s | No | No |
| Sound | Generated with every clip | Generated with every clip | Generated with the picture |
| Seed | Yes | Yes | No |
| Price per second | $0.081 / $0.117 / $0.171 (720p / 1080p / 1440p) | $0.108 / $0.153 (720p / 1080p) | $0.02 / $0.04 / $0.045 (480p / 720p / 1080p) |
| 10 s at 1080p | $1.17 | $1.53 | $0.45 |
Source: the request rules and prices of the three models on apimodels.app, 2026-11-20. Lightricks positions Fast for speed and cost and Pro for fidelity and stable motion; we have not published our own quality comparison.
Why use LTX 2.5 Fast here: apimodels.app vs the LTX API, fal.ai and self-hosting
LTX 2.5 Fast access routes compared on capabilities only, cell by cell, including the rows where the other route is stronger.
| apimodels.app | LTX API (Lightricks) | fal.ai | Self-hosting the open weights | |
|---|---|---|---|---|
| Text, image and audio to video | One model id; the mode follows your inputs | Separate endpoints per mode | Separate endpoints per mode | Separate pipelines or ComfyUI workflows |
| Highest resolution (Fast) | 1440p | 4K | 4K (2160p) | 4K with the spatial upscaler |
| Last frame | No | Yes (last_frame_uri) | Yes (end_image_url) | Yes (first-and-last-frame workflow) |
| Silent output | No, sound is always generated | Yes (generate_audio false) | Yes (generate_audio false) | Not checked |
| Automatic duration | No, fixed steps | Yes (duration null) | Yes ("auto") | Yes (optional duration predictor) |
| Text-to-video aspect ratios | 16:9, 9:16, 1:1, 3:2, 2:3 | 16:9, 9:16, 1:1, 3:2, 2:3 | 16:9, 9:16 | Any size divisible by 32 |
| Audio-to-video: resolution and fps choice | 720p / 1080p / 1440p, 24–50 fps | Yes | Not offered (no resolution or fps field) | Yes (audio-to-video workflow) |
| Audio-to-video: trim to a set length | Yes, duration 2–20 s | No, length follows the audio | No duration field | Not checked |
| Audio-to-video on Pro | No | Yes | Yes | Not applicable |
| Upload size limits | Image 10 MB, audio 50 MB | Image 15 MB, audio 32 MB | Not published on the model page | Your hardware |
| Same key for other models | Yes, image, video, audio and LLM models on one balance | LTX models only | Yes, fal hosts many model APIs | Not applicable |
| Fine-tuning / LoRA | No | Not checked | Yes, own weights on fal Serverless | Yes, trainable checkpoint on Hugging Face |
| Own GPUs needed | No | No | No | Yes |
Checked on 2026-11-20 from each provider’s own pages: Lightricks’ LTX-2.5 model page, image-to-video and audio-to-video API references and input-format limits on docs.ltx.io; fal.ai’s LTX-2.5 model pages (llms.txt) for each mode and tier; the LTX-2.5 model card on Hugging Face. "Not checked" means we did not find it on those pages.
Choose apimodels.app when you want one model id for three modes, audio-to-video at 720p–1440p with a set length, and the same key for other models. Choose Lightricks’ API or fal.ai when you need 4K, a last frame, silent output or Pro audio-to-video. Self-host when you want to fine-tune and can run the GPUs.
LTX 2.5 Fast limits and pricing, in numbers
What one LTX 2.5 Fast clip costs, and every limit a request has to stay inside.
Billing
LTX 2.5 Fast bills per second of output. Mode and frame rate do not change the rate. Audio-to-video is billed on the output length to 0.01 s; without duration it follows the audio, capped at 20 s (10 s at 1440p or 48 / 50 fps).
Inputs
An LTX 2.5 Fast prompt is 1–5,000 characters and required in every mode. One first frame: JPG, PNG or WEBP up to 10 MB. One audio file: MP3, WAV, M4A, AAC, OGG or FLAC, at least 2 s, up to 50 MB, and not shorter than the duration you set.
Not available here
LTX 2.5 Fast here has no 4K, last frame, reference images, video input, automatic duration or silent output. Requests with these fields return a 400 before a task is created, so nothing is charged.
Failures and results
LTX 2.5 Fast success is not 100%: content review or a busy upstream can fail a task, and failed tasks are not charged. Result files are kept for 30 days.
How long a clip takes
In our tests on 20 November 2026, LTX 2.5 Fast returned 6-second 720p clips in 27 to 73 seconds and a 4.8-second audio-to-video clip in 86 seconds, queue included. Five runs, not a guarantee: busy hours and longer or higher-resolution clips take longer.
| Resolution | Per second | 6 s | 10 s | 20 s | Lightricks list price |
|---|---|---|---|---|---|
| 720p | $0.081 | $0.486 | $0.81 | $1.62 | $0.09/s |
| 1080p | $0.117 | $0.702 | $1.17 | $2.34 | $0.13/s |
| 1440p | $0.171 | $1.026 | $1.71 | Not available (max 10 s) | $0.19/s |
Sources
Checked on 2026-11-20. Statements marked as Lightricks’ come from these pages and are the vendor’s claims, not our tests; prices on this page are apimodels.app prices.
- LTX API pricing (Lightricks): ltx-2-5-fast and ltx-2-5-pro per-second rates, Fast vs Pro ↗
- LTX-2.5 model page (Lightricks): multi-shot, prompt adherence, auto duration ↗
- LTX-2.5 support matrix (docs.ltx.io): resolutions, fps, durations ↗
- LTX audio-to-video API reference (docs.ltx.io) ↗
- LTX prompting guide (docs.ltx.io) ↗
- LTX-2.5 model card and license (Hugging Face) ↗
- LTX release notes: LTX-2.5, 11 Aug 2026 ↗
- fal.ai: LTX-2.5 Fast text-to-video model page ↗
- fal.ai: LTX-2.5 Fast audio-to-video model page ↗
About LTX 2.5 Fast
LTX 2.5 Fast is a Video Generation API provided by Lightricks. LTX 2.5 Fast is the speed-and-cost tier of LTX-2.5, the open-weights video and audio model Lightricks released on 11 August 2026, and every clip it makes comes with sound generated alongside the picture. On APIMODELS it is the model id ltx-2.5-fast on POST /v1/video/generations, from $0.081 per second at 720p, which is 10% below Lightricks' own API list price, and failed tasks are not charged. One model id covers three modes and picks the mode from what you send: a prompt alone is text-to-video; a prompt plus one first-frame image in image_url is image-to-video; a prompt plus an audio track in audio_url is audio-to-video, where your audio becomes the soundtrack and the picture is generated to match it, with or without a first frame. Resolution is 720p, 1080p (the default) or 1440p, at 24, 25, 48 or 50 fps; frame rate does not change the price. Text-to-video and image-to-video run 6 to 20 seconds in even steps at 720p and 1080p with 24 or 25 fps, and 6, 8 or 10 seconds at 1440p or at 48 or 50 fps. Audio-to-video follows the length of your audio unless you set duration, which can be 2 to 20 seconds with decimals (up to 10 at 1440p or 48 / 50 fps). Text-to-video takes aspect_ratio 16:9, 9:16, 1:1, 3:2 or 2:3; image-to-video keeps the shape of the image, and audio-to-video without an image is 16:9. seed (0 to 2147483647) is optional. Billing is per second of output: $0.081 at 720p, $0.117 at 1080p and $0.171 at 1440p, so a 6-second 720p clip is $0.486, a 10-second 1080p clip $1.17, a 20-second 1080p clip $2.34 and a 10-second 1440p clip $1.71. Audio-to-video is billed on the output length to 0.01 s: 12.5 seconds at 1080p is $1.4625. Lightricks lists ltx-2-5-fast at $0.09, $0.13 and $0.19 per second on its API pricing page and describes Fast as built for speed and cost, for iteration, previews and high-volume generation, with Pro for final renders. Lightricks also says LTX-2.5 adds native multi-shot scenes that keep characters, lighting and voice consistent across cuts, a new diffusion video decoder and better prompt adherence; those are the vendor's claims, not our tests. Limits on APIMODELS: no 4K, no last frame, no reference images and no video input; one first frame in JPG, PNG or WEBP up to 10 MB; audio in MP3, WAV, M4A, AAC, OGG or FLAC, at least 2 seconds and up to 50 MB; prompts up to 5,000 characters. Not every request succeeds: content review or a busy upstream can fail a task, and failed tasks are not charged. The API is asynchronous: the create call returns a task id, then you poll GET /v1/video/generations?task_id= until the state is completed, or pass callback_url. Result files are kept for 30 days. Through APIMODELS platform, you can access this model via a unified API with transparent pay-as-you-go pricing. Current pricing: 720p · per second: $0.081, 1080p · per second: $0.117, 1440p · per second: $0.171, 720p · 6s: $0.486, 1080p · 10s: $1.17, 1080p · 20s: $2.34, 1440p · 10s: $1.71, Audio-to-video: $same rate, billed on output length to 0.01 s.
Related model: LTX 2.5 Pro — the fidelity tier of the same model for final shots — 720p / 1080p, 6, 8 or 10 seconds, $0.108/s and $0.153/s, no audio-to-video
Related model: LTX-2.3 — the previous LTX — 480p / 720p / 1080p in 16:9 or 9:16, 5-20 seconds, from $0.02/s
Related model: MiniMax H3 — when one shot needs several references — up to 9 images, 3 videos and 3 audio clips, 5-15 seconds at 768P or native 2K
Related model: Seedance 2.5 — when the clip must run past 20 seconds — 4-30 seconds at 480p / 720p, with video editing and extension
Key Features
- Three Modes, One Model ID -- A prompt alone is text-to-video; add image_url for image-to-video from a first frame; add audio_url for audio-to-video. The mode follows what you send
- Audio-to-Video -- Your MP3, WAV, M4A, AAC, OGG or FLAC track (2 s or longer) becomes the soundtrack and the picture is generated to match it. Length follows the audio, or set 2–20 s; billed to 0.01 s
- 720p, 1080p or 1440p -- At 24, 25, 48 or 50 fps. Text-to-video takes 16:9, 9:16, 1:1, 3:2 or 2:3; image-to-video keeps the shape of your first frame
- 6–20 Seconds -- Even seconds from 6 to 20 at 720p / 1080p with 24 or 25 fps; 6, 8 or 10 s at 1440p or 48 / 50 fps. Sound is generated with every clip
- 10% Below the ltx.io List Price -- $0.081 / $0.117 / $0.171 per second at 720p / 1080p / 1440p, against $0.09 / $0.13 / $0.19 on Lightricks’ API pricing page. Frame rate does not change the price
- Limits, Stated Plainly -- No 4K, last frame, reference images or video input here. Content review or a busy upstream can fail a task; failed tasks are not charged
Use Cases
Marketing Videos
Quickly generate brand promotion videos for ad campaigns and social media marketing.
Social Media Content
Create compelling short-form video content for platforms like TikTok, Instagram, and YouTube.
Product Demos
Generate product feature demonstrations and tutorials to improve user conversion.
Educational Content
Produce course explanations, knowledge explainers, and training videos at an affordable price.
Why APIMODELS
- Unified API -- One API key to access all models, no need to register on multiple platforms
- Transparent Pricing -- Pay-as-you-go with per-call pricing published on every model page, ideal for indie developers and startups
- Instant Access -- Start right after signup; pay by card, PayPal, Alipay or USDT
- Full Documentation -- Detailed API docs with code examples in cURL, Python, and Node.js
Frequently Asked Questions
What is the difference between APIMODELS and the official Lightricks API?
APIMODELS offers the same LTX 2.5 Fast model through our aggregation platform. We provide a unified API interface so you do not need separate accounts for each provider - one API key to access all models.
What is LTX 2.5 Fast, and how do I use the LTX 2.5 API?
LTX 2.5 Fast is the speed-and-cost tier of LTX-2.5, the 22B-parameter video and audio model (the checkpoints on Hugging Face are named ltx-2.5-22b) that Lightricks released on 11 August 2026. On apimodels.app you call it with one API key: POST https://api.apimodels.app/v1/video/generations with {"model": "ltx-2.5-fast", "prompt": "...", "resolution": "1080p", "duration": 10}. The call returns a taskId; poll GET /v1/video/generations?task_id= until state is completed and read data.resultUrls[0], or pass callback_url to be notified. The same model id also does image-to-video (add image_url) and audio-to-video (add audio_url). You do not need a Lightricks account. Copy-paste curl examples for every mode are on /docs/ltx-2-5.
LTX release notes: LTX-2.5 API, 11 Aug 2026 ↗How much does the LTX 2.5 Fast API cost?
$0.081 per second at 720p, $0.117 at 1080p and $0.171 at 1440p, billed on the length of the video you get back. That is 10% below Lightricks' own list price for ltx-2-5-fast ($0.09, $0.13 and $0.19 per second). So 6 seconds at 720p is $0.486, 10 seconds at 1080p is $1.17 and 20 seconds at 1080p is $2.34. The mode and the frame rate do not change the rate; audio-to-video is billed to 0.01 s of output, so 12.5 seconds at 1080p is $1.4625. Failed tasks are not charged. There is no free tier: the $0.10 trial credit on new accounts is less than the shortest clip.
LTX API pricing (Lightricks) ↗LTX 2.5 vs MiniMax H3: what is the difference?
They take different inputs. LTX 2.5 Fast on apimodels.app makes 6–20 second clips (even seconds) at 720p, 1080p or 1440p and 24 to 50 fps, from a prompt, one first-frame image, or an audio track that the picture is built around; it takes no reference images or reference video. MiniMax H3 here makes 5–15 second clips at 768P or native 2K from a prompt plus up to 9 reference images, 3 reference videos and 3 reference audio clips, in aspect ratios from 21:9 to 9:16 plus adaptive. Both generate sound with the picture. LTX 2.5 fits work that starts from a script, a single frame or a voice track; MiniMax H3 fits shots where several references must hold a character, a motion or a voice. In our same-prompt test (three 6-second clips each), LTX 2.5 Fast finished in 27 to 64 seconds and MiniMax H3 at 768P in 4.1 to 4.8 minutes, while H3 kept the staging, the shot order and the product closer to the prompt every time.
Our same-prompt test: LTX 2.5 vs MiniMax H3 (apimodels.app blog) ↗How do I write a good LTX 2.5 prompt? Is there a prompt guide?
Yes, Lightricks publishes one. Its core advice: write a single flowing paragraph in the present tense, about 4 to 8 sentences for one shot, that covers the shot type, the scene and light, the action in order, the characters, the camera move and the sound, with spoken lines in quotation marks. Example: "Close-up of an old fisherman mending a net on a wooden pier at dawn, low warm sun on his face. He pulls the cord tight, looks up and says, 'Storm's coming.' The camera slowly pushes in while gulls cry and waves slap the posts." For several shots in one clip, describe each cut in prose ("a hard cut to a wide shot of...") and keep it to 2–4 shots. Lightricks also warns that on-screen text and chaotic physics are still uneven and that prompts written for other models should be rewritten, not pasted. Prompts can be up to 5,000 characters here.
LTX prompting guide (Lightricks docs) ↗How does LTX 2.5 image-to-video work?
Add one image to the request as image_url (a public URL or a base64 data URI; JPG, PNG or WEBP up to 10 MB) and describe the motion in the prompt. The clip opens on that image and keeps its aspect ratio, so leave aspect_ratio out; sending a different ratio returns a 400. Lengths and prices are the same as text-to-video: 6–20 seconds in even steps at 720p or 1080p with 24 or 25 fps, 6, 8 or 10 seconds at 1440p or 48 / 50 fps, and 10 seconds at 1080p costs $1.17. Only a first frame is supported: there is no last frame and no reference images, and a second image is rejected. Lightricks' guide suggests keeping image-to-video to one continuous take unless you deliberately describe a cut away from the opening image.
Can LTX 2.5 do lip sync or audio-to-video?
Yes, on ltx-2.5-fast. Send audio_url with a speech, singing or music track (MP3, WAV, M4A, AAC, OGG or FLAC, at least 2 seconds, up to 50 MB) and a prompt: the track becomes the soundtrack and the picture is generated to match it. Add a portrait as image_url to make that person talk or sing. Without duration the clip follows the audio, up to 20 seconds (10 at 1440p or 48 / 50 fps); billing is the output length to 0.01 s, so 12.5 seconds at 1080p is $1.4625. In our own test, a 4.8-second voice line plus a portrait came back in 86 seconds as a 4.8-second clip of the person saying it, with the mouth following the words, for $0.39. Results vary with the audio and the face, so try a short clip first. Audio-to-video is not available on ltx-2.5-pro here, and without an image it is always 16:9.
Our test clips: LTX 2.5 vs Seedance 2.0 (apimodels.app blog) ↗LTX 2.5 vs LTX 2.3: what changed, and is 2.5 faster?
On apimodels.app they are separate model ids with different limits. LTX 2.5 Fast adds 1440p, a choice of 24, 25, 48 or 50 fps, aspect ratios 1:1, 3:2 and 2:3, audio-to-video and seed, and runs 6–20 seconds from $0.081 per second. LTX 2.3 here offers 480p, 720p or 1080p in 16:9 or 9:16, 5–15 seconds for text-to-video or 5–20 for image-to-video, from $0.02 per second. Lightricks says LTX-2.5 adds native multi-shot scenes, a new diffusion video decoder for sharper faces and on-screen text, a Gemma 4 12B text encoder that holds complex prompts together, and a faster distilled model; those are the vendor's claims. Through our API, 6-second 720p LTX 2.5 Fast clips took 27 to 73 seconds from request to finished file in our November 2026 tests; we did not time LTX 2.3 alongside it. Stay on LTX 2.3 for 480p drafts from $0.02 per second; move to 2.5 for 1440p, square or 3:2 framing, or a clip driven by an audio track.
LTX-2.5 model card (Hugging Face, Lightricks) ↗LTX 2.5 Fast or LTX 2.5 Pro: which should I use?
Use Fast for iteration, longer clips and audio: it reaches 20 seconds and 1440p, it is the only tier with audio-to-video here, and it starts at $0.081 per second. Use Pro when the shot is final and fits in 10 seconds at 720p or 1080p; Lightricks tunes Pro for visual fidelity and motion stability, at $0.108 (720p) or $0.153 (1080p) per second. At 1080p a 10-second clip is $1.17 on Fast and $1.53 on Pro. The request is identical apart from the model string, so you can settle the prompt on Fast and render the chosen shot on ltx-2.5-pro.
Is LTX 2.5 free or open source? Can I get it on Hugging Face?
The weights are open: Lightricks publishes LTX-2.5 on Hugging Face (Lightricks/LTX-2.5) under the LTX-2.x Community License, which allows commercial use at no cost for organisations under $10M in annual revenue; above that a paid licence is required. Running it yourself means your own GPUs and setup (ComfyUI or Lightricks' Python pipelines). The hosted API is paid: on apimodels.app ltx-2.5-fast starts at $0.081 per second, with no subscription and no charge for failed tasks. The $0.10 trial credit on new accounts does not cover the shortest clip ($0.486).
LTX-2.5 on Hugging Face (license and weights) ↗What is the max length and max resolution of LTX 2.5 Fast? Does it do 4K or first frame + last frame?
On apimodels.app LTX 2.5 Fast goes up to 20 seconds and 1440p: 20 seconds needs 720p or 1080p at 24 or 25 fps, while 1440p or 48 / 50 fps caps a clip at 10 seconds. There is no 4K (Lightricks' own API has it), no last frame, no reference images and no reference video; a request with any of them returns a 400 and is not charged. One first frame (JPG, PNG or WEBP, up to 10 MB), one audio track (MP3, WAV, M4A, AAC, OGG or FLAC, at least 2 seconds, up to 50 MB) and prompts up to 5,000 characters. Durations come in fixed steps, not any whole second: even seconds from 6 to 20, or 6, 8 or 10 at 1440p or 48 / 50 fps. aspect_ratio applies to text-to-video only. Success is not 100%: content review or a busy upstream can fail a task; failed tasks are not charged, so adjust the input and retry. Result files are deleted after 30 days.
LTX 2.5 vs Wan 3.0 or Seedance 2.5: which fits my job?
It depends on the input you start from and the length you need. On apimodels.app LTX 2.5 Fast keeps your audio track as the soundtrack and builds the picture around it, and it goes to 1440p, but clips stop at 20 seconds and there are no reference images. Wan 3.0 Video here runs up to 30 seconds at 480P, 720P or 1080P, with a first and last frame or up to 10 reference images. Seedance 2.5 here runs 4–30 seconds at 480p or 720p, takes up to 30 reference images, 10 videos and 10 audio clips, and can edit or extend an existing clip. All three generate sound with the picture. These are spec differences only; we have not run a side-by-side quality test, so try the same prompt on each model page.
What can LTX 2.5 Fast do?
LTX 2.5 Fast supports: Text to Video, Image to Video (First Frame), Audio to Video, 720p / 1080p / 1440p, 6-20s, 24 / 25 / 48 / 50 fps, Native Audio, Seed, From $0.081/s. See the APIMODELS docs for full parameters and call examples.
Can I use the LTX 2.5 Fast API from any country, including mainland China?
Yes. APIMODELS does not restrict access by country: the site and the API answer requests from anywhere, and apimodels.app and api.apimodels.app are reachable directly from mainland China without a proxy. You call LTX 2.5 Fast with one APIMODELS key, so you do not need a separate account with the model's developer.
What payment methods are supported? Can I pay with PayPal or crypto?
Yes to both. You can top up with a card (Visa, Mastercard and others, through Stripe), PayPal, Alipay, or USDT on TRON (TRC20) or BNB Smart Chain (BEP20). Card, PayPal and Alipay top-ups are credited immediately; USDT is credited once the transfer confirms on-chain. Auto top-up is optional, and a $50 top-up adds a $1 bonus ($3 on a $100 top-up). The balance is in USD and billed per use, with no subscription.
Related models
LTX 2.5 prompt examples
See all 89 →Prompts shared by their authors — copy and adapt them. Each one credits its author and links back to the original post.
Woman Gazes Toward Glitching Monitor
A close-up of a woman, his eyes fixed straight ahead, narrowed and locked on something in the far distance directly in front of her, her bre…
by @ivanfioravanti

Chocolate Poured Over Cream and Berries
Style: high-resolution food photography - close-up - The chocolate is poured over the whipped cream and berries. The chocolate flows slowly,…
by hichuicomfy

Waterfall Cabin With Orange Autumn Trees
A stylized illustration depicts a serene natural scene with a waterfall cascading down a rocky cliff into a blue river. Two trees with vibra…
by Esteban099009
Need prompts?
We curate copy-ready prompt libraries — every entry shows its full text and a sample result, ready to adapt.
Comparisons
How this model stacks up against the alternatives on results, price and features.