Alpha Gen
- New
/v2/video-alpha-genendpoint (async) to separate the foreground of a video from its background. Send a video and get back a grayscale alpha matte video with the same frames, white for foreground and black for transparent, plus a ProRes 4444 copy of your video with that matte as its alpha channel, ready to composite.
Video Upscale
- New
/v2/video-upscaleendpoint (async) to upscale a video to 1080p, 4K or 8K. The output keeps the input’s aspect ratio and audio, and uses its frame rate rounded to a whole number.
HDR output color space
/v2/video-to-video-hdrnow takes an optionaloutput_colorspace(acescg,acescct,srgb_linear, oraces2065-1) to choose the color space of the exported EXR sequence (exr_frames_url).- Requests without
output_colorspaceare unchanged and default toacescg.
SDR-to-HDR upgrade
- Upgraded
/v2/video-to-video-hdrwith an HLG video output alongside the EXR sequence and support for longer input videos.
V1 video generation retired
- As announced in the September 24 retirement notice,
/v1/text-to-video,/v1/image-to-video,/v1/audio-to-video,/v1/retake, and/v1/extendare no longer available. Requests to these endpoints return an error. - Use the corresponding V2 endpoints instead. The media upload endpoint remains
/v1/upload.
LTX-2.5 square and 3:2 / 2:3 aspect ratios
ltx-2-5-fastandltx-2-5-pronow generate 1:1, 3:2, and 2:3 video at every resolution from 720p to 4K, on text-to-video, image-to-video, and audio-to-video.- Pass one of the new sizes, such as
1080x1080or1620x1080, inresolution. Each is priced the same as the 16:9 size with the same short side.
See the LTX-2.5 resolutions and Pricing.
Video Reframe placement
- Reframe now takes an optional
placement, to choose where the source video lands on the output frame. Pass an anchor such as{ "anchor": "top" }to fit and align it, or a rectangle such as{ "x": 0, "y": 0.35, "width": 1 }to set its position and size exactly. - Requests without
placementare unchanged: the source is centered at the largest size that fits.
V1 video generation retirement date
- After October 26, 2026 at 11:59 PM UTC, requests to
/v1/text-to-video,/v1/image-to-video,/v1/audio-to-video,/v1/retake, and/v1/extendwill no longer work. - Migrate to the corresponding V2 endpoints before then. The media upload endpoint remains
/v1/upload.
Video Reframe
- Reframe now picks its own model, so you no longer need to send
model. If you still send it, it’s ignored. Existing requests keep working.
V1 video generation deprecation
- V1 video generation endpoints with V2 replacements are being deprecated:
/v1/text-to-video,/v1/image-to-video,/v1/audio-to-video,/v1/retake, and/v1/extend. - Move to the corresponding V2 endpoints to submit background jobs, poll for completion, and download the result. Your authentication, generation request parameters, and pricing remain unchanged.
- The media upload endpoint remains
/v1/upload.
Audio-to-video generation params
- Audio-to-video now supports
fps(default 24),last_frame_uri(requiresimage_uri), andcamera_motion— matching text-to-video and image-to-video. Available on v2/audio-to-video and v1/audio-to-video.
Audio-to-video resolution parity
- Audio-to-video now supports the same resolution tiers as text-to-video and image-to-video:
ltx-2-5-fastup to 4K,ltx-2-5-proup to 1080p, andltx-2-3-proup to 4K. - Maximum input audio length depends on model and resolution tier — for example,
ltx-2-3-proaccepts up to 20 seconds at 720p/1080p and up to 10 seconds at 1440p/4K. See the API reference for the full table. - New resolution tiers are billed at the rates shown on Pricing.
LTX-2 removal
- As announced in the July 2 LTX-2 deprecation notice,
ltx-2-fastandltx-2-proare no longer available. Requests that specify these models return an error. - Migrate to LTX-2.3 (
ltx-2-3-fast,ltx-2-3-pro) or LTX-2.5 (ltx-2-5-fast,ltx-2-5-pro).
See Supported Models and Pricing.
LTX-2.5
- New LTX-2.5 models for Text-to-Video, Image-to-Video, and Audio-to-Video, in portrait and landscape:
ltx-2-5-fastup to 4K, andltx-2-5-proat 720p and 1080p. Audio-to-Video generates at 1080p on both. - Both models accept camera motion,
last_frame_urion Image-to-Video, and"duration": nullfor automatic duration.
Automatic duration
- Text-to-Video and Image-to-Video now accept
"duration": nullonltx-2-5-fast, letting the model choose the clip length from your prompt.
See Automatic duration and Pricing.
720p video generation
- Text-to-Video and Image-to-Video now support a 720p output tier on LTX 2.3 (
1280x720landscape and720x1280portrait).
See Pricing and the supported models.
New API domain: api.ltx.io
- The LTX API is now available at
https://api.ltx.io— docs and code samples use it going forward. - Existing integrations on
https://api.ltx.videokeep working unchanged, so no migration is required.
Video Reframe
- New
/v2/video-to-video-reframeendpoint (async) to reframe a video to a new resolution, filling in the missing areas with generated content (720p and 1080p; 1:1, 4:5, 5:4, 9:16, 16:9).
LTX-2 deprecation
- LTX-2 (
ltx-2-fast,ltx-2-pro) will be deprecated on July 15, 2026. From that date, LTX-2 requests are automatically served by LTX-2.3 with no change to LTX-2 pricing. On August 15, 2026, LTX-2 will be removed and those requests return an error. - Migrate to LTX-2.3 (
ltx-2-3-fast,ltx-2-3-pro) before August 15, 2026.