AI Clipping ComfyUI Nodes

July 21, 2026 · View on GitHub

A free, open-source Opus Clip alternative for ComfyUI. Turn long-form videos, podcasts, interviews, lectures, and livestreams into ranked viral-ready short clips for YouTube Shorts, TikTok, and Instagram Reels — directly inside your ComfyUI workflow. Single managed API call: transcription, virality ranking, dedupe, and face-tracked auto-crop all run server-side via muapi.ai. No local Whisper, no local LLM, no GPU.

Searching for an Opus Clip alternative, a Klap.app alternative, or a Vizard / 2short.ai alternative that you can run inside ComfyUI? You're in the right place. This pack ships AI clipping, AI highlight extraction, automatic vertical reframing, face tracking, and viral-clip ranking — all as native ComfyUI nodes.

ai-clipping-generator — Ready-made Next.js SaaS for AI video clipping — no ComfyUI needed
AI-Youtube-Shorts-Generator — Full YouTube Shorts generator with virality ranking and auto-crop

Why This vs. Opus Clip / Klap / Vizard?

	This (AI Clipping nodes)	Opus Clip / Klap / Vizard
Native ComfyUI integration	✅ — clip, then chain into upscale / lipsync / VHS / etc.	❌ — closed web app
Pricing	Pay-as-you-go via muapi.ai (no subscription)	Monthly subscription tiers
Self-hostable / scriptable	✅ — open-source nodes + open API	❌
Aspect ratios	9:16 / 1:1 / 4:5	9:16 / 1:1 / 4:5
Face-tracked vertical crop	✅	✅
Viral-score per clip	✅ (0–100 + hook line + reason)	✅
Coordinates-only mode	✅ — render with your own ffmpeg / VHS pipeline	❌

If you're shipping shorts at scale and want clipping to be one node in a bigger ComfyUI graph (intro/outro overlays, watermark removal, AI b-roll, captions, lipsync, etc.), this is the cleanest path.

What is AI Clipping?

AI Clipping converts a long-form video (podcast, interview, lecture, vlog, livestream, …) into ranked short-form clips suitable for TikTok / Reels / YouTube Shorts. The server handles the entire pipeline:

Whisper transcription — audio → timestamped segments
Highlight ranking through a virality framework — hook moments, emotional peaks, opinion bombs, revelation moments, conflict, quotable lines, story peaks, practical value
Overlap dedupe — collapses near-duplicate candidates by score
Top-N selection — keeps the highest-scoring N
Face-tracked vertical auto-crop — renders each clip at the requested aspect ratio

Each clip ships with a viral score (0–100), an opening hook line, and a one-sentence "why it works" reason — all surfaced as a JSON output you can pass to downstream nodes.

Reference implementation (open source): SamurAIGPT/AI-Youtube-Shorts-Generator Underlying API: https://muapi.ai/playground/ai-clipping

Nodes

Node	Description
🔑 AI Clipping API Key	Set your key once — wire to all nodes
✂️ AI Clipping	Long video → ranked vertical short clips (full pipeline)
✂️ AI Clipping Pick Clip	Pick the Nth clip from the highlights JSON for fan-out
✂️ AI Clipping Save Video	Download a clip URL → disk + ComfyUI IMAGE frames

Installation

Via ComfyUI Manager (recommended)

Open ComfyUI Manager → Install via Git URL
Paste: https://github.com/Anil-matcha/ai-clipping-comfyui
Restart ComfyUI

Manual

cd ComfyUI/custom_nodes
git clone https://github.com/Anil-matcha/ai-clipping-comfyui
pip install -r ai-clipping-comfyui/requirements.txt

Quick Start

Sign up at muapi.ai and grab a key from Dashboard → API Keys → Create Key.
Right-click the ComfyUI canvas → Add Node → ✂️ AI Clipping.
Add a 🔑 AI Clipping API Key node, paste your key, and wire its output into the AI Clipping node.
Paste a hosted video URL (or a local file path), set num_clips, hit Queue Prompt.

Tip: If you use the MuAPI CLI, run muapi auth configure --api-key YOUR_KEY once and every node will pick the key up automatically — no need to paste it anywhere.

Node Reference

🔑 AI Clipping API Key

Set your muapi.ai API key once and wire the output to all AI Clipping nodes. Alternatively, leave every api_key field blank — nodes auto-read from ~/.muapi/config.json if you've authenticated via the CLI.

✂️ AI Clipping

Run the full clipping pipeline server-side and return a ranked list of vertical short clips.

Field	Values	Default	Notes
`video_url`	URL string	—	Hosted MP4 / MOV URL of the source video
`video_path`	Local path	—	Alternative to `video_url` — file is auto-uploaded to muapi
`num_clips`	1 – 20	`3`	How many highlights to extract
`aspect_ratio`	`9:16` / `1:1` / `4:5`	`9:16`	`9:16` for Shorts/TikTok/Reels, `1:1` for IG Feed, `4:5` portrait
`return_coordinates_only`	bool	`false`	Return only highlight time ranges, skip cropping
`api_key`	Optional — leave blank if using the API Key node or CLI config	—

Provide either video_url or video_path (not both). When return_coordinates_only=true, the response carries start_time/end_time per highlight but no clip_url — render the clips yourself with ffmpeg or any downstream node.

Outputs

first_clip_url (STRING) — URL of the top-ranked clip (most useful to pipe directly into Save Video)
first_frame (IMAGE) — first frame of the top clip, ready for previews / further processing
highlights_json (STRING) — full JSON array of every clip, with metadata (title, score, hook_sentence, virality_reason, start_time, end_time, clip_url). Pass this to Pick Clip to fan out.
request_id (STRING) — for re-fetching the result later
clip_count (INT) — how many clips actually came back

Endpoint: POST /api/v1/ai-clipping

✂️ AI Clipping Pick Clip

Pick the Nth clip from the highlights_json returned by the main node. Use this to fan out — e.g. send clip #1 to one Save Video, clip #2 to another, etc.

Field	Values	Default	Notes
`highlights_json`	JSON array	—	Wire from `AI Clipping` → `highlights_json`
`index`	1 – 50	`1`	1-based rank (#1 = top-scoring clip)
`load_first_frame`	bool	`true`	Download the clip and return its first frame

Outputs: clip_url · first_frame (IMAGE) · title · hook_sentence · start_time · end_time · score

Out-of-range indices clamp to the available range. If the input list is empty, returns blanks.

✂️ AI Clipping Save Video

Downloads a clip URL to ComfyUI's output folder and returns all frames as an IMAGE tensor for use with other nodes (preview, VHS, upscale, etc.).

Field	Description
`video_url`	URL returned by AI Clipping or Pick Clip
`save_subfolder`	Subfolder under `ComfyUI/output/` (default: `ai_clipping`)
`filename_prefix`	Filename prefix (default: `clip`)
`frame_load_cap`	Optional max frames returned (`0` = all)
`skip_first_frames`	Skip N frames from the start
`select_every_nth`	Stride for frame selection

Outputs: frames (IMAGE) · filepath (STRING) · frame_count (INT)

Example Workflows

Load AiClipping_Example.json via File → Load in ComfyUI.

File	Description
`AiClipping_Example.json`	Long video URL → top clip → save to disk + preview frame
`AiClipping_FanOut_Example.json`	Same pipeline plus Pick Clip nodes for #2 and #3 saved separately

Single best clip:

[🔑 API Key] ───────────────────────────────────────────────────────┐
                                                                     ↓
(video URL) → [✂️ AI Clipping] → first_clip_url → [✂️ Save Video] → frames → [Preview Image]

Fan-out — pick clips #1, #2, #3 separately:

                                            ┌→ [✂️ Pick Clip #1] → [✂️ Save Video]
[✂️ AI Clipping] → highlights_json ─────────┼→ [✂️ Pick Clip #2] → [✂️ Save Video]
                                            └→ [✂️ Pick Clip #3] → [✂️ Save Video]

Aspect Ratio Picker

Platform	Ratio	Sweet-spot duration
TikTok / Reels / YouTube Shorts	`9:16`	30–75s
Instagram Feed	`1:1`	15–45s
Pinterest / portrait	`4:5`	30–60s

Default to 9:16 unless the platform is specified.

Output Schema (`highlights_json`)

[
  {
    "title": "The one mistake that cost me \$50K",
    "start_time": 124.3,
    "end_time": 187.6,
    "score": 92,
    "hook_sentence": "Nobody talks about this, but it killed my first startup...",
    "virality_reason": "Opens with a number + regret, peaks on a contrarian lesson",
    "clip_url": "https://.../short_1.mp4"
  }
]

When return_coordinates_only=true, each entry has start_time/end_time but no clip_url — render locally.

API

This node pack uses the muapi.ai API under the hood:

AI Clipping: POST https://api.muapi.ai/api/v1/ai-clipping
Poll: GET https://api.muapi.ai/api/v1/predictions/{request_id}/result
Upload: POST https://api.muapi.ai/api/v1/upload_file

Authentication is a single x-api-key header — no session tokens required.

The submit-then-poll flow:

POST /api/v1/ai-clipping   →  { "request_id": "abc123" }
GET  /api/v1/predictions/abc123/result
                           →  { "status": "processing" }       (keep polling)
                           →  { "status": "completed",
                                "shorts": [ {...}, {...}, ... ] }

Status values: queued, pending, processing, completed, failed.

Requirements

Python ≥ 3.8
requests ≥ 2.28 · Pillow ≥ 9.0 · numpy ≥ 1.23 · torch ≥ 2.0 · opencv-python ≥ 4.7

Want More Models?

This repo is focused on AI Clipping only. If you need access to 100+ models — Kling, Veo3, Flux, HiDream, GPT-image-1.5, Imagen4, Wan, lipsync, audio, image enhancement and more — check out the full MuAPI ComfyUI node pack:

SamurAIGPT/muapi-comfyui — ComfyUI nodes for every muapi.ai model in one place.

License

_{Keywords: Opus Clip alternative, Opus Clip ComfyUI, Opus.pro alternative, Klap alternative, Klap.app alternative, Vizard alternative, 2short.ai alternative, free Opus Clip, open-source Opus Clip, AI video clipper, AI clipping, AI clipper for ComfyUI, ComfyUI shorts generator, ComfyUI TikTok generator, ComfyUI Reels generator, viral clip generator, podcast clipper, podcast to shorts, long video to shorts, video highlight extraction, AI highlight extractor, automatic vertical reframing, face-tracked vertical crop, autocrop video, viral score, hook detection, ComfyUI video editing, muapi.ai, YouTube Shorts maker, TikTok clip maker, Instagram Reels maker.}