ComfyUI Node
LTX 2.3 Generate All-in-One
Single-pass native LTXV generation with prompts, image guides, sampling, video decode, and optional audio output.
LTX 2.3 Generate All-in-One
- model
- clip
- vae
- start_images
- audio
- audio_vae
- images
- audio
◄positive_prompt►
◄negative_prompt►
◄width768►
◄height512►
◄fps24.00►
◄num_frames97►
◄timing_modeframe►
◄resize_modecontain►
◄duplicate_policyerror►
◄pad_color0,0,0►
◄img_compression35►
◄global_strength1.00►
◄lock_start_framesfalse►
◄lock_end_framefalse►
◄start_images_strength0.85►
◄seed0►
◄steps30►
◄cfg1.0►
◄sampler_nameeuler_cfg_pp►
◄max_shift2.05►
◄base_shift0.95►
◄stretchtrue►
◄terminal0.10►
◄sigma_modeltx_scheduler►
◄manual_sigmas1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0►
◄audio_modepassthrough►
◄guides_json{"version":1,"guides":[]}►
CategoryLTX 2.3
Inputs (33)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | Native LTXV or LTXV AV model used for single-pass sampling. | |
| clip | CLIP | Text encoder used to encode the positive and negative prompts. | |
| vae | VAE | LTXV video VAE used for guide encoding and final video decode. | |
| positive_prompt | STRING | Positive text prompt encoded inside this generation node. | |
| negative_prompt | STRING | Negative text prompt encoded inside this generation node. | |
| width | INT | 76864–16384 | Generated video width. Must be divisible by 32 after rounding down. |
| height | INT | 51264–16384 | Generated video height. Must be divisible by 32 after rounding down. |
| fps | FLOAT | 24.001–240 | Frames per second used when timing_mode is seconds. |
| num_frames | INT | 971–16384 | Pixel frame count used for timing, negative frame positions, and internally-created empty latents. Native LTXV lengths must be 8*n + 1, for example 97, 105, 113. |
| timing_mode | COMBO | frame | Interpret manual guide positions as frame indexes or seconds. |
| resize_mode | COMBO | contain | How guide images are resized before VAE encoding. contain/pad preserves aspect ratio with padding. |
| duplicate_policy | COMBO | error | How to handle manual guide images that resolve to the same frame. |
| pad_color | STRING | 0,0,0 | RGB padding color for contain/pad resize mode. Accepts r,g,b or #rrggbb. |
| img_compression | INT | 350–100 | Native LTXV image compression applied before guide encoding. Set 0 to disable. |
| global_strength | FLOAT | 1.000–1 | Multiplier applied to every manual guide strength and start sequence strength. |
| lock_start_frames | BOOLEAN | false | When enabled, frame 0 guides and start_images are written into the beginning video latent instead of only appended as guide references. VAE-level lock, not pixel-perfect copy. |
| lock_end_frame | BOOLEAN | false | When enabled, a manual guide resolving to the final frame is written into the final video latent instead of only appended as a guide reference. VAE-level lock, not pixel-perfect copy. |
| start_images_strength | FLOAT | 0.850–1 | Strength for the optional start image sequence before global_strength is applied. |
| seed | INT | 00–18446744073709550000 | Noise seed for generation. |
| steps | INT | 301–10000 | Number of sampling steps for the native LTXV scheduler. |
| cfg | FLOAT | 1.00–100 | Classifier-free guidance scale. |
| sampler_name | COMBO | euler_cfg_pp | Sampler algorithm used by ComfyUI's custom sampler. |
| max_shift | FLOAT | 2.050–100 | Native LTXV model sampling and scheduler max shift. |
| base_shift | FLOAT | 0.950–100 | Native LTXV model sampling and scheduler base shift. |
| stretch | BOOLEAN | true | Stretch scheduler sigmas to the terminal value. |
| terminal | FLOAT | 0.100–0.99 | Terminal scheduler sigma value when stretch is enabled. |
| sigma_mode | COMBO | ltx_scheduler | Use native LTX scheduler sigmas or a pasted ManualSigmas-style schedule. |
| manual_sigmas | STRING | 1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0 | Comma/space separated sigma values used when sigma_mode is manual. Requires at least 2 values. |
| audio_mode | COMBO | passthrough | passthrough outputs connected audio or silence. native_av samples audio/video latents with an LTXV AV model and audio_vae. |
| guides_json | STRING | {"version":1,"guides":[]} | Hidden serialized guide data used by the custom UI and saved in workflows. |
| start_imagesopt | IMAGE | Optional IMAGE batch from a video source. Applied as a native multi-frame guide starting at frame 0. | |
| audioopt | AUDIO | Optional external audio. In passthrough mode it is trimmed/padded and output; in native_av mode it is encoded as a locked audio latent. | |
| audio_vaeopt | VAE | Required for native_av audio mode. Use the native LTXV Audio VAE Loader. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| audio | AUDIO | — |