LTX 2.3 / editVideo

LTX 2.3's video-editing node, running on Civitai's cloud: you feed it a source video and a prompt, and it regenerates the clip along the lines of the original - like an img2img for footage, done server-side. It's one of the six LTX 2.3 nodes in civitai/civitai-comfy-nodes, and the one whose dials you'll actually touch is the canny pair, because this edit path is structurally guided by an edge map.

The edit-vs-v2v distinction matters here. The sibling videoToVideo node also takes a source video, but it keeps the whole clip as conditioning with a guide_strength defaulting to 0.2 - gentle restyling. This editVideo node defaults guide_strength to 1.0 and adds canny_low_threshold / canny_high_threshold (defaults 0.4 / 0.8): the source footage is passed through a Canny edge detector and the edges are what hold the composition together while the model re-renders content. Canny is the classic structural-control trick from image ControlNet land (covered in the KB's controlnet doc), lifted into video. Lower thresholds keep more edges, which holds structure harder; raise them to give the model more room to re-imagine.

How it works

Standard orchestration: videoGen workflow with the LTX 2.3 engine, submit to the Civitai Orchestration API, long-poll, download the MP4 as a VIDEO. Your source_video uploads to Civitai's servers and the canny edge map is computed as part of the hosted Comfy workflow. The workflow_id and raw_json outputs let you inspect what actually ran when an edit surprises you.

Inputs and outputs that matter

• prompt (required) - the edit: style change, restyle, subject swaps.
• source_video (required) - the VIDEO socket with your footage.
• negative_prompt - what to avoid.
• duration - tooltip: 3 or 5 seconds.
• width / height - default 1280×720; divisible by 32.
• fps - default 24; steps - default 20 (8–50); guidance_scale - default 4; generate_audio - default true.
• model - 22b-dev / 22b-distilled.
• loras - CIVITAI_LORAS socket.
• diffusion_model - transformer override for community fine-tunes.
• quantity - 1–10 variants per job, distinct seeds.
• canny_low_threshold / canny_high_threshold - the edge map's thresholds (0–1).
• guide_strength - 0–1, default 1.0; how hard the edited output sticks to the source structure.
• api_config - auth override.

Outputs: video, workflow_id, raw_json.

Installing it

ComfyUI Manager → Custom Nodes Manager → "Civitai Comfy Nodes" (publisher civitai) → Install → restart, or clone + pip install -r civitai-comfy-nodes/requirements.txt. Only requests; no models.

Common issues

• Auth & Buzz. Civitai account with Buzz; CIVITAI_API_TOKEN for headless installs.
• Over-editing or stuck-in-place. guide_strength near 1.0 with low canny thresholds keeps the source nearly frame-identical; drop guide_strength and raise thresholds when you want real change.
• Canny artifact bleed. Edge maps carry high-frequency detail. If the output is full of jittery line-like artifacts, raise canny_low_threshold to suppress weak edges.
• Upload time. The source video uploads to Civitai; keep sources short and compressed.
• Early preview. Node behavior can change without notice; pin the version for anything you rely on.
• Cost. Editing is a full cloud video job; read the per-run cost shown on the node.