RuneXX

/

LTX-2.3-Workflows

like 293

Just a simple log for those that have already downloaded workflows, and curious to what might have been changed since.
For considerations to update and re-download or not ;-)

1st of april:

• UPDATED: V2V Foley workflow updated to have multiple video outputs to choose from (raw input video + audio added OR LTX re-created video+ audio that can have slight alterations fitting the prompt)

30th of march:

• NEW: V2V - ReTake - recreate a section within your input video with new content. Re-shoot/re-create an area of any input video.
• UPDATED: I2V & T2V - ID-LoRA - consistent voice from 5 second reference audio. A small update with higher strength to the ID-Lora node and some tweaks for higher consistency.

28th of march:

• NEW: I2V & T2V - "Talking Avatar" or "Digital Human" kind of workflows where you can prompt consistent dialog across videos with Qwen TTS or Fish Audio S2 Pro

25th of march:

• NEW: I2V & T2V - ID-LoRA - consistent voice across video generations with a 5 second reference audio

22nd of march:

• NEW: V2V - Just Talk - Prompt lip-synced voice and sounds to any silent video. With optional extend video part of same workflow.

18th of march:

• UPDATED: V2V - Add Sound to any video improvements for smoother results and usage
• UPDATED: V2V - Extend-Any-Video workflow fixed a routing bug, and added some improvements.

17th of march:

• NEW: I2V & T2V Dev Model Full Steps workflow (the same workflow can be used both for single pass and 2-pass with upscaler)
  This workflow has lots of "bells & whistles" so there are quite a few settings to play around with. Steps, video cfg, and audio cfg... feel free to experiment to squeeze out max ;-)
  Being full steps, quite a bit slow. So for most users the other workflows with distilled model is probably more practical to use.

16th of march:

• NEW: Improved long length I2V+T2V with custom audio video workflow with better consistency, and much lighter and simpler.
  This new workflow is single pass. And the "older" 2-pass long length workflows with be updated with same improvements asap ;-)

14th of march:

• UPDATED: V2V Foley workflow updated to work better.

13th of march:

• NEW: I2V & T2V Workflow for long videos, with custom audio (ideal for long shot of singer, dialogs etc).
  One workflow has multiple groups extending the video, while the other has one group that runs in a loop (uses "ReTake" and masking for consistency)
• NEW: V2V Foley - Add sound to any video. Input any silent video and prompt for sounds you want added to the video; sound fx, music, even narrator voice.
• NEW: V2V Extend any video. Take an input video and make it longer. Use prompt to steer how the extended part will be.
  Uses LTX "Retake"/masking to make extension be consistent with input video for motion, looks, and voice.

10th of march:

• UPDATED: Added a simple switch to I2V T2V custom audio workflow to switch between vocals only (for improved lip-sync) or full audio (to include sound fx, music etc)
  (the final video has full audio no matter what, but the switch determines what LTX hear when generating the video)

8th of march:

• NEW: From user request: First-Last Frame & First-Last-Middle-Frame with custom audio (if lack of lip-sync transcribe the audio dialog in the prompt)
• NEW: Single Pass workflows added (one sampler pass, direct render no downscale + upscale)

7th of march:

• NEW: First Middle Last & First Last Frame workflow added
• UPDATED: All workflows updated to use euler_cfg & new sigma values at the sampler to align with LTX own settings
  (And thanks to raffetazarius for testing a bunch samplers to compare outputs)

6th of march:

• added new tiny vae for sampler preview to all the workflows