muapi-ai-clipping
Turn a long video into N viral-ready short clips with a single managed API call.
Paste into Claude, ChatGPT or Cursor.
Read the source194 lines
| 1 | |
| 2 | slug muapi-ai-clipping |
| 3 | name muapi-ai-clipping |
| 4 | version "1.0.0" |
| 5 | description Turn a long video into N viral-ready short clips with a single managed API call. Wraps muapi.ai's `/ai-clipping` endpoint, which handles transcription, highlight ranking through a virality framework (hook / emotional peak / opinion bomb / revelation / conflict / quotable / story peak / practical value), overlap dedupe, and vertical face-tracking auto-crop server-side. No local Whisper, no local LLM, no GPU. |
| 6 | acceptLicenseTerms true |
| 7 | |
| 8 | |
| 9 | # AI Clipping |
| 10 | |
| 11 | **One API call: long video in → ranked vertical short clips out.** |
| 12 | |
| 13 | Each clip ships with a viral score (0–100), an opening hook line, a one-sentence "why it works" reason, and a hosted mp4 URL. |
| 14 | |
| 15 | Underlying API: https://muapi.ai/playground/ai-clipping |
| 16 | Reference implementation (open source): https://github.com/SamurAIGPT/AI-Youtube-Shorts-Generator |
| 17 | |
| 18 | |
| 19 | |
| 20 | ## When to Use |
| 21 | |
| 22 | Auto-clip a podcast, interview, lecture, vlog, or stream into TikTok / Reels / Shorts. |
| 23 | Extract the best 30–75s moments from any hosted video URL. |
| 24 | Get face-tracked vertical (9:16), square (1:1), or portrait (4:5) crops without running ffmpeg locally. |
| 25 | |
| 26 | If you only need raw timestamps for your own renderer, set `--coords-only` to skip cropping and just get the highlight ranges. |
| 27 | |
| 28 | |
| 29 | |
| 30 | ## Agent Execution Protocol |
| 31 | |
| 32 | ### Step 1 — Collect Inputs |
| 33 | |
| 34 | | Input | Required | Default | Notes | |
| 35 | |:---|:---|:---|:---| |
| 36 | | `--video` | yes | — | Hosted mp4 URL, or local file path (auto-uploaded), or YouTube URL (if backend supports it) | |
| 37 | | `--num-clips` | no | `3` | Number of highlights to extract | |
| 38 | | `--aspect-ratio` | no | `9:16` | `9:16` \| `1:1` \| `4:5` | |
| 39 | | `--coords-only` | no | off | Return just the highlight time ranges, skip cropping | |
| 40 | |
| 41 | If the user gave only a video URL, run with defaults — don't block on questions. |
| 42 | |
| 43 | |
| 44 | |
| 45 | ### Step 2 — Verify Prerequisites |
| 46 | |
| 47 | `muapi-cli` installed and authed (`muapi auth configure`) |
| 48 | `MUAPI_API_KEY` available (env var or `muapi auth status` passes) |
| 49 | |
| 50 | That's it. No `ffmpeg`, no Python, no Whisper install, no LLM keys. Everything runs server-side. |
| 51 | |
| 52 | |
| 53 | |
| 54 | ### Step 3 — Run the Skill |
| 55 | |
| 56 | |
| 57 | bash library/edit/ai-clipping/scripts/run-ai-clipping.sh \ |
| 58 | --video "https://example.com/podcast.mp4" \ |
| 59 | --num-clips 5 \ |
| 60 | --aspect-ratio 9:16 \ |
| 61 | --view |
| 62 | |
| 63 | |
| 64 | The script: |
| 65 | Resolves `--video` to a hosted URL (uploads local files via `muapi upload file` if needed). |
| 66 | Calls `muapi edit clipping` with the supported parameters. |
| 67 | Polls until the job is done (or returns the `request_id` immediately under `--async`). |
| 68 | Prints a ranked summary and, if `--output-json` is set, writes the full result. |
| 69 | |
| 70 | |
| 71 | |
| 72 | ## What Happens Server-Side |
| 73 | |
| 74 | The `/ai-clipping` endpoint internally runs the full pipeline so the agent doesn't have to: |
| 75 | |
| 76 | **Transcribe** with Whisper. |
| 77 | **Classify content type** (podcast / interview / tutorial / vlog / lecture / monologue). |
| 78 | **Rank highlights** through the virality framework: |
| 79 | **Hook moments** — strong opening line that stops the scroll |
| 80 | **Emotional peaks** — laughter, anger, vulnerability, awe |
| 81 | **Opinion bombs** — spicy, contrarian, debate-bait takes |
| 82 | **Revelation moments** — "wait, what?" reframes |
| 83 | **Conflict** — disagreement, tension, callouts |
| 84 | **Quotable lines** — tight, screenshot-worthy phrasing |
| 85 | **Story peaks** — climax of a narrative arc |
| 86 | **Practical value** — actionable insight a viewer will save |
| 87 | **Dedupe** overlapping candidates by score. |
| 88 | **Top-N select** and **face-track auto-crop** to the requested aspect ratio. |
| 89 | |
| 90 | This is why the skill is small: the heavy lifting is on the API. |
| 91 | |
| 92 | |
| 93 | |
| 94 | ## Quick Invocation Patterns |
| 95 | |
| 96 | **Defaults — three 9:16 clips:** |
| 97 | |
| 98 | bash run-ai-clipping.sh --video "https://example.com/long.mp4" |
| 99 | |
| 100 | |
| 101 | **Podcast — more clips, view in player:** |
| 102 | |
| 103 | bash run-ai-clipping.sh --video "<URL>" --num-clips 8 --view |
| 104 | |
| 105 | |
| 106 | **Square clips for Instagram feed:** |
| 107 | |
| 108 | bash run-ai-clipping.sh --video "<URL>" --aspect-ratio 1:1 --num-clips 3 |
| 109 | |
| 110 | |
| 111 | **Just the timestamps (build your own renderer):** |
| 112 | |
| 113 | bash run-ai-clipping.sh --video "<URL>" --coords-only --output-json result.json |
| 114 | |
| 115 | |
| 116 | **Async submit (returns request_id, poll later):** |
| 117 | |
| 118 | REQUEST_ID=$(bash run-ai-clipping.sh --video "<URL>" --async --output-json - | jq -r '.request_id') |
| 119 | muapi predict wait "$REQUEST_ID" --download ./outputs |
| 120 | |
| 121 | |
| 122 | **Local file:** |
| 123 | |
| 124 | bash run-ai-clipping.sh --video ./recording.mp4 --num-clips 5 --view |
| 125 | |
| 126 | |
| 127 | **Batch — `urls.txt` with one URL per line:** |
| 128 | |
| 129 | xargs -a urls.txt -I{} bash run-ai-clipping.sh --video "{}" |
| 130 | |
| 131 | |
| 132 | |
| 133 | |
| 134 | ## Aspect Ratio Picker |
| 135 | |
| 136 | | Platform | Ratio | Sweet-spot duration | |
| 137 | |:---|:---|:---| |
| 138 | | TikTok / Reels / YouTube Shorts | `9:16` | 30–75s | |
| 139 | | Instagram Feed | `1:1` | 15–45s | |
| 140 | | Pinterest / portrait | `4:5` | 30–60s | |
| 141 | |
| 142 | Default to `9:16` unless the platform is specified. |
| 143 | |
| 144 | |
| 145 | |
| 146 | ## Output Schema |
| 147 | |
| 148 | |
| 149 | { |
| 150 | "source_video_url": "...", |
| 151 | "shorts": [ |
| 152 | { |
| 153 | "title": "The one mistake that cost me $50K", |
| 154 | "start_time": 124.3, |
| 155 | "end_time": 187.6, |
| 156 | "score": 92, |
| 157 | "hook_sentence": "Nobody talks about this, but it killed my first startup...", |
| 158 | "virality_reason": "Opens with a number + regret, peaks on a contrarian lesson", |
| 159 | "clip_url": "https://.../short_1.mp4" |
| 160 | } |
| 161 | ] |
| 162 | } |
| 163 | |
| 164 | |
| 165 | When `--coords-only` is set, each entry has `start_time`/`end_time` but no `clip_url` — render locally with ffmpeg. |
| 166 | |
| 167 | When reporting back to the user, surface for each clip: rank, score, time range, title, hook, and clip URL. |
| 168 | |
| 169 | |
| 170 | |
| 171 | ## Common Mistakes to Avoid |
| 172 | |
| 173 | **Wrong aspect ratio for the platform** — Shorts / TikTok / Reels are `9:16`. Default to that. |
| 174 | **Padding to hit `num_clips`** — if the API returns fewer survivors than requested, return what you have. Don't pretend. |
| 175 | **Re-running on a 404'd clip URL** — the same `request_id` can be re-fetched with `muapi predict wait <id>` rather than re-clipping. |
| 176 | **Trying to tune Whisper / chunk size / LLM prompts** — those knobs aren't exposed; the endpoint handles them. |
| 177 | |
| 178 | |
| 179 | |
| 180 | ## Failure Modes |
| 181 | |
| 182 | **API key missing or rejected** — surface the exact error; never fabricate a key. |
| 183 | **Job timed out** — bump poll timeout (`--poll-timeout`) and retry. |
| 184 | **Source URL not reachable from the backend** — upload locally with `muapi upload file <path>` first, then pass the returned URL. |
| 185 | **Fewer clips returned than requested** — the source had fewer rankable highlights. Return what came back with a note. |
| 186 | |
| 187 | |
| 188 | |
| 189 | ## Done Criteria |
| 190 | |
| 191 | The skill is done when: |
| 192 | `result.shorts` has up to `num_clips` entries, each with a working `clip_url` (or `start_time`/`end_time` under `--coords-only`). |
| 193 | The user has been shown the ranked list (score, time range, title, hook, URL). |
| 194 | If `--output-json` was set, the file exists and parses. |