dreamina-3-0/pro/text-to-video
Turn text into detailed cinematic scenes with Dreamina 3.0 precision.
LTX 2.5 is Lightricks' open-weights audio-video model for generating synchronized picture and sound from text, images, or audio. This page runs the Image-to-Video Fast tool: animate a still into a clip with optional native audio in a speed-focused mode built for quick drafts. Sibling RunComfy pages cover text-to-video and audio-to-video Fast modes in the same LTX 2.5 family.
Creators pick LTX 2.5 when they need open weights plus hosted Fast tools. Across tasks, LTX 2.5 keeps identity and sound more aligned while you iterate. Many teams start on this Image-to-Video Fast page, then move to other LTX 2.5 task pages without changing creative direction.
| LTX 2.5 advantage | What it means for you |
|---|---|
| Native multi-shot continuity | Direct connected scenes in one pass so character look, set, lighting, voice, and grade stay aligned across cuts. |
| Adaptive detail where it counts | Rendering budget shifts with scene complexity, keeping faces, textures, and on-screen type sharper when needed. |
| Updated video decoder | Cleaner motion, fewer reconstruction artifacts, and more stable detail versus earlier LTX generations. |
| Stronger prompt hold | A Gemma 4 12B text encoder plus prompt enhancer helps LTX 2.5 retain multi-character, camera, and lighting instructions. |
| Fast iteration tier | Choose duration auto or fixed lengths, and scale from 720p through 4K (2160p) with audio included in the rate. |
LTX 2.5 covers still animation, text-led drafts, and audio-timed clips across sibling Fast pages; this tool focuses on image-to-video.
auto or 6–20s), resolution (720p–2160p), aspect ratio (auto, 16:9, 9:16), FPS, audio on/off, and optional camera motion.Controls on this LTX 2.5 Image-to-Video Fast page:
| Parameter | Required | Type | Default | Range / Options | How to choose |
|---|---|---|---|---|---|
image_url* | Yes (*) | Image URL | Example still | JPG, JPEG, PNG, WEBP, GIF, AVIF | Use a sharp, well-lit still; composition largely locks the framing LTX 2.5 will animate. |
end_image_url | No | Image URL | — | Same formats | Add only for a start→end morph; leave empty for open-ended motion from the first frame. |
prompt* | Yes (*) | String | Example prompt | Up to 5,000 characters | Cover subject action, camera, lighting, style, and audio sources in one coherent brief. |
duration | No | String | auto | auto, 6, 8, 10, 12, 14, 16, 18, 20 | Use auto to let LTX 2.5 infer length; pick fixed seconds when timing must match an edit. |
resolution | No | String | 1080p | 720p, 1080p, 1440p, 2160p | Draft at 720p; move up for hero placements and large crops. |
aspect_ratio | No | String | auto | auto, 16:9, 9:16 | Prefer auto to follow the still; force 16:9 or 9:16 for platform delivery. |
fps | No | Integer | 25 | 24, 25, 48, 50 | Match editorial cadence (24/25) or smoother motion (48/50). |
generate_audio | No | Boolean | true | true / false | Keep on for synced ambience, SFX, or dialogue; off for silent picture plates. |
camera_motion | No | String | — | dolly_in, dolly_out, dolly_left, dolly_right, jib_up, jib_down, static, focus_shift | Optional macro move when you want a prescribed camera path on top of the prompt. |
Billing for LTX 2.5 on this page is per second of generated video. LTX 2.5 Image-to-Video Fast includes native audio at every resolution.
| Resolution | Price per second | 6s | 10s | 20s |
|---|---|---|---|---|
| 720p | $0.10 | $0.60 | $1.00 | $2.00 |
| 1080p | $0.14 | $0.84 | $1.40 | $2.80 |
| 1440p | $0.21 | $1.26 | $2.10 | $4.20 |
| 4K (2160p) | $0.33 | $1.98 | $3.30 | $6.60 |
For batches, total ≈ duration × per-second rate × output count. When duration is auto, LTX 2.5 billing follows the actual generated length.
Reusable structure for LTX 2.5 (keep one main action per clip):
[who/what is on screen] + [action over time] + [camera] + [lighting & environment] + [look/grade] + [audio sources] + [ending beat]
generate_audio is on.end_image_url when the finish pose or layout must land precisely.Weak prompt
> Make the photo cinematic with nice camera movement and sound.
Improved prompt for LTX 2.5
> A woman in a charcoal wool coat stands under a clear umbrella at a rain-slick Tokyo crosswalk at night. She lowers the umbrella, steps into the crosswalk as a yellow taxi glides through the mid-ground, and neon kanji signs ripple across the wet asphalt. Locked-off wide shot, fine rain streaks, soft reflections. Audio: light rain, distant traffic hum, one soft tire hiss on water.
Use this LTX 2.5 family guide; exact limits depend on the endpoint. Based on publicly available information.
| Model | Standout | Typical fit |
|---|---|---|
| LTX 2.5 | Open-weights AV model with multi-shot continuity, Fast modes, and resolutions through 4K | Synced audio-video from stills, text, or audio with open weights available |
| LTX 2 | Earlier Lightricks AV generation with Fast/Pro tiers | Pipelines that do not yet need 2.5 continuity upgrades |
| Kling 3.0 | Multi-shot prompting with strong lip-sync options | Scripted character ads and dialogue-led spots |
| Seedance 2.0 / 2.5 | Dense reference mixing across image, video, and audio | Brand work that must lock identity from many references |
| MiniMax H3 | Multimodal family with stereo audio and reference workflows | Prompt-first drafts that may later add Omni-style references |
Choose LTX 2.5 when open weights and Fast cloud endpoints both matter. What sets LTX 2.5 apart is multi-shot continuity with native audio in one pass. Run the same brief across models before locking a production path.
Published LTX 2.5 weights also support local fine-tuning when hosted drafts are not enough.
Turn text into detailed cinematic scenes with Dreamina 3.0 precision.
Fast, low-cost multi-shot AI video model with native audio and references.
Turn stills into cinematic motion clips with camera and audio control.
HappyHorse 1.0 Reference to Video fuses up to 9 reference images and a prompt into a coherent multi-character clip with stable identity.
Prompt-driven song creation with 44.1 kHz WAV control and section editing
Interpolates start-end frames with refined motion control presets
LTX 2.5 animates a still into a short video with optional synchronized audio. On this page you use the Image-to-Video Fast tool for rapid drafts, while sibling pages cover text-to-video and audio-to-video Fast modes in the same model family.
LTX 2.5 can generate connected scenes in one pass so character appearance, environment, lighting, voice, and visual style stay more consistent across cuts. That helps campaign animatics and multi-beat spots that previously needed separate single-shot generations.
You can generate at 720p, 1080p, 1440p, or 4K (2160p), with duration set to auto or fixed lengths of 6–20 seconds. FPS options include 24, 25, 48, and 50, and aspect ratio can follow the still (auto) or use 16:9 or 9:16.
Yes. Provide a start image (required) and optionally an end image when you want a guided transition between two compositions. Leave the end image empty for open-ended motion from the first frame only.
Native audio can be generated with the picture when generate_audio is enabled, and audio is included in the per-second rate at every resolution. Turn audio off if you need a silent plate for later sound design.
Prompts can be up to 5,000 characters. Start and optional end images accept common formats such as JPG, PNG, WEBP, GIF, and AVIF. Check the current RunComfy parameter panel for any additional upload size limits.
Yes. Prototype in the RunComfy Web UI, then call the same LTX 2.5 Image-to-Video Fast model via the HTTP API with matching parameters for automation and production workflows.
Billing is per second of generated video: $0.10 at 720p, $0.14 at 1080p, $0.21 at 1440p, and $0.33 at 4K (2160p). When duration is auto, the charge follows the actual generated length. Generations consume USD/credits on RunComfy; new users typically receive a free trial amount.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.
