Quick Answer
Accounts that post 3-5 short clips a day grow 4-6x faster than those posting once. The blocker was never ideas, it was edit time. AI drops it to about 20 minutes per clip.
What you get
- Produce a captioned, reframed short clip in 15-25 minutes instead of 1-2 hours
- Post 3-5 clips a day without a full-time video editor
- Turn one 30-minute video into 8-12 postable clips per session
- Build faceless clips from scratch with AI voice and avatar when you do not want to film
- Run the required stack for $35/mo, or the full repurpose-and-faceless stack for $110/mo
Step-by-Step Workflow
6 steps · 2 hours to set up · 20 min per clip ongoing
Workflow at a glance
6 steps · 2 hours setup
Pick pipeline
Write hooks
ClaudeAuto-clip (A)
Opus ClipFaceless build (B)
ElevenLabs
HeyGenCaption + polish
Caption + post
ClaudePick pipeline
Write hooks
Auto-clip (A)
Faceless build (B)
Caption + polish
Caption + post
- 1
Pick your pipeline: repurpose long video or build faceless
Decide up front. If you have source video (a podcast, a talk, a webinar, or a webcam recording), run Pipeline A - it is cheaper per clip and performs better because it is a real person. If you do not film yourself, run Pipeline B - script plus AI voice plus b-roll or an AI avatar.
Do not mix them mid-session. Batch by pipeline: one session of repurposing, or one session of faceless clips. Context-switching between the two is where the 20-minute-per-clip target slips.
5 minOutput: Chosen pipeline and a batch of source material or topics - 2
Write hooks first - the clip lives or dies in 2 seconds
Before any editing, generate hooks. Prompt Claude: 'Give me 10 opening-line hooks for a short video about [topic] for [audience]. Each under 8 words, designed to stop a scroll. Mix formats: a bold claim, a question, a mistake-to-avoid, a number, and a curiosity gap. No clickbait I cannot pay off.'
For Pipeline B, follow with: 'Now write a 30-second script for the strongest hook - hook, one payload point, one closing line that earns a follow. Spoken-word cadence, short sentences.'
The hook is the highest-leverage element in short-form. A great clip with a weak hook gets 200 views; an average clip with a strong hook gets 20,000.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 10 min for a batchOutput: A batch of hooks, plus scripts for faceless clipsTools: ClaudeTip: Front-load the payload. Say the most interesting thing in the first line, not last. Short-form audiences do not wait for a build-up; retention is decided before the 3-second mark.
- 3
Pipeline A: auto-clip and reframe the long video with Opus Clip
Upload your long video to Opus Clip. It scans for the most sharable moments, cuts them into 30-60 second clips, reframes to 9:16 with speaker tracking so the face stays centered, adds captions, and scores each clip for virality.
Review the clips it picks - keep the ones where the hook lands in the first 2 seconds, discard the rest. Swap in a stronger opening line from your Claude hook batch where a clip has a weak start. One 30-minute source video typically yields 8-12 keepers.
Opus Clip - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 15 min for a batch of clipsOutput: 8-12 reframed, captioned clips from one long videoTools: Opus Clip - 4
Pipeline B: generate voice and visuals for the faceless clip
Take the script from step 2. Generate the voiceover in ElevenLabs (pick one consistent voice and reuse it - a recognizable voice builds a brand). For visuals, either pair the voiceover with relevant b-roll in CapCut, or generate an AI avatar reading the script in HeyGen for a talking-head look without filming.
Keep faceless clips tight: 20-40 seconds, one idea, fast cuts every 2-3 seconds to hold attention. Faceless works, but it needs pace to compensate for the missing human presence.
ElevenLabs - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 15 min per clipOutput: A finished faceless clip with AI voice and visualsTools: ElevenLabs, HeyGen, CapCut - 5
Add premium captions and a final polish
Captions are non-negotiable - most short-form is watched on mute. Opus Clip and CapCut both auto-caption; Submagic adds animated word-by-word captions, keyword highlights, emojis, and auto b-roll that measurably lift retention.
Run the clip through Submagic (or style captions in CapCut if you are keeping the stack lean), then do a 2-minute polish: trim any dead air at the start, punch in the hook, add a single sound effect or zoom on the key line. Do not over-edit - past a point, polish stops adding views.
5-10 min per clipOutput: Final captioned, polished vertical clipTools: Submagic, CapCutTip: Set your caption style once (font, color, position, animation) and reuse it on every clip. A consistent caption look becomes part of your brand and stops you re-deciding it every time.
- 6
Write the on-platform caption and post natively
Generate the post caption with Claude: 'Write a scroll-stopping caption for this clip for [platform]. First line is a hook that works even if the video does not autoplay. Add a soft CTA and 3-5 relevant, non-spammy hashtags.'
Upload natively to each platform (TikTok, Reels, Shorts) rather than cross-posting a watermarked file - platforms suppress reach on clips that carry another app's watermark or logo. Post at a consistent daily cadence; short-form rewards volume and consistency over any single perfect clip.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 5 min per clipOutput: Posted clip with native caption and hashtagsTools: ClaudeTip: Remove watermarks before posting. TikTok, Reels, and Shorts all quietly downrank clips that visibly came from another app - export clean from Opus Clip or CapCut.
Short-form is a volume game. The accounts that win post multiple clips a day, and the only thing stopping most people is that editing a single clip - reframing to vertical, adding captions, finding the hook - used to take an hour or two. AI collapsed that. This workflow gives you two pipelines: Pipeline A repurposes long video (a podcast, a webinar, a talk) into a batch of clips with auto-reframing and captions, and Pipeline B builds faceless clips from scratch using an AI script, AI voice, and optionally an AI avatar. The required stack is $35/mo (Opus Clip plus Claude). Add premium captions, manual polish, AI voice, and an avatar and it tops out at $110/mo.
The two ways to make short-form with AI in 2026
There are exactly two efficient pipelines, and picking the wrong one wastes time. Pipeline A (repurpose) is for anyone already creating long video - it uses auto-clipping to find the sharable 30-60 second moments, reframe them vertical with speaker tracking, and burn in captions. One 30-minute source video yields 8-12 clips, so your cost per clip approaches zero once the long video exists. Pipeline B (faceless from scratch) is for people who do not film themselves - it scripts a hook and payload with an LLM, generates a voiceover, and pairs it with stock b-roll or an AI avatar. Pipeline A is far cheaper per clip and gets higher engagement because it is a real human; use it whenever you have source video. Pipeline B is the fallback when you do not want to be on camera. The universal truth across both: the first 1-2 seconds (the hook) decides whether the clip is watched at all, and captions are non-negotiable because most short-form is watched on mute.
Stack cost breakdown
Public list prices as of July 2026. Optional tools are marked in the notes.
| Tool | Plan | Monthly cost | Notes |
|---|---|---|---|
| Opus Clip | Starter | $15/mo | Required. Auto-clip, reframe to vertical, caption. |
| Claude | Pro | $20/mo | Required. Hooks, scripts, caption copy, titles. |
| Submagic | Starter | $20/mo | Optional. Premium animated captions, b-roll, sound effects. |
| CapCut | Pro | $20/mo | Optional. Manual polish, transitions, templates (free tier works). |
| ElevenLabs | Starter | $6/mo | Optional. AI voiceover for faceless clips. |
| HeyGen | Creator | $29/mo | Optional. AI avatar for faceless talking-head clips. |
| Total | $35 - $110/mo($35 required, $110 with optional tools) | ||
Email me this stack as a checklist
Every tool, the plan to pick, and the monthly cost - in your inbox.
Real usage
What people actually run
No usage reports yet - be the first to share what you run. Tell us your real stack, your actual monthly cost, and any tools you swapped.
The prompts that make the clips work
Short-form is won on the hook and lost on the mute. These two prompts cover both - a hook batch you draw from all week, and a script built for spoken cadence.
Hook batch prompt (Claude)
Give me 10 opening-line hooks for a short video about [topic] for [audience]. Rules: each under 8 words, designed to stop a scroll in the first second. Mix formats - a bold claim, a question, a common mistake, a specific number, a curiosity gap, a contrarian take. No hook I cannot pay off in the clip. Rank them by scroll-stopping power.
Note: Generate a batch weekly, not per clip. Pulling from a ready bank of hooks is what keeps per-clip time at 20 minutes instead of 45.
Faceless script prompt (Claude)
Write a 30-second faceless short-form script on [topic] for [audience]. Structure: hook (first line, under 8 words, front-load the payoff), one clear payload point with a specific example, one closing line that earns a follow. Spoken-word cadence - short sentences, no complex clauses, easy to say out loud. Mark where a b-roll cut or on-screen text should land.
Note: The 'easy to say out loud' constraint matters - LLM prose is often written, not spoken, and it sounds robotic when an AI voice reads it. This prompt fixes the cadence.
Adjust for Your Situation
If you already run a podcast or YouTube channel
Pipeline A is nearly free for you - the source video already exists. Run every long video through Opus Clip and you get 8-12 clips per episode at effectively $15/mo total. Skip the faceless tools entirely. This is the highest ROI version of the workflow.
If you refuse to be on camera
Commit to Pipeline B. Make ElevenLabs and either CapCut (b-roll) or HeyGen (avatar) required. Pick one AI voice and one visual style and hold them constant so the account still feels like a brand. Expect slightly lower engagement than face content and compensate with faster pacing and higher volume.
If you are optimizing for one platform
Match the clip length and style to the platform's sweet spot: TikTok rewards 21-34 second clips with fast hooks, Reels favors 7-15 seconds for reach and 30-60 for saves, Shorts tolerates up to 60 and rewards clean retention. Do not post one identical clip everywhere - trim to each platform's length in CapCut.
Swap options
Drop-in substitutions if a tool does not fit your budget or stack. These trade cost or effort for the recommended setup.
| Swap out | Use instead | When |
|---|---|---|
| Opus Clip | Vizard or Klap | You want more free processing minutes or a different clip-selection model |
| Submagic | CapCut auto-captions | The free CapCut caption styles are close enough and you want to skip a subscription |
| HeyGen | Film yourself on a phone | You are comfortable on camera - real faces still outperform AI avatars on engagement |
| Claude | ChatGPT | You already pay for it; the hook and script prompts work in either |
Common Pitfalls
- Posting clips with another app's watermark. TikTok, Reels, and Shorts downrank them. Always export clean.
- Weak or slow hooks. Retention is decided in the first 2 seconds. Front-load the payoff; never build up to it.
- No captions. Most short-form is watched on mute. A clip without captions loses the majority of its audience instantly.
- Over-editing a single clip. Past a point, polish adds no views. Ship at 'good' and post the next one - volume beats perfection.
- Mixing pipelines in one session. Batch repurposing and faceless separately, or the context-switching blows past the 20-minute-per-clip target.
- Using AI avatars when you could film yourself. Real faces outperform avatars on engagement. Only go faceless if you genuinely will not film.
Frequently Asked Questions
Which is better, repurposing long video or building faceless clips?
Do I need Submagic if Opus Clip and CapCut already caption?
How many clips can I realistically make per hour?
Will AI-generated clips get suppressed by the algorithm?
Can I run this for free?
What is the single biggest lever on views?
How we built this workflow
Prices verified July 2026 from each vendor's pricing page: Opus Clip Starter $15/mo, Claude Pro $20/mo, Submagic Starter $20/mo, CapCut Pro $19.99/mo (shown as $20), ElevenLabs Starter $6/mo, HeyGen Creator $29/mo. The 4-6x growth figure for high-frequency posting and the 15-25 minute per-clip range reflect common short-form operator patterns; niche, hook quality, and platform drive most of the variance.
Last updated July 7, 2026; prices verified at publication.