Script in. Published Short out. No camera.

Find a Short in your niche that already works. You are not copying the footage. You are reverse-engineering the structure so your version follows beats that are proven to hold attention.

1. 1.Pull the audio and transcribe it with word timestamps
2. 2.Grab one frame every half second and lay them out as a contact sheet
3. 3.Write down the timeline: when the hook card ends, when captions start, when the punch-ins hit
4. 4.Name the sections. A typical talking-head Short has 5 or 6: hook card, split screen, punch-in, mockup, payoff, open loop

The format is the asset. The footage is disposable.

Hook in the first 3 seconds. Value in the middle. Payoff at the end. Open loop or CTA last. Under 120 words for a 40-second Short.

1. 1.Match your script to the section timeline from step 1
2. 2.One idea per Short. If it needs two, it is two Shorts
3. 3.Read it out loud once. If you stumble, the voice model will too

One clean 2 to 3 minute recording of you talking is enough to clone your voice in 11 Labs. The clone is the easy part. Making it sound like you on a good day is the setting.

1. 1.Generate with the v3 model, not multilingual v2
2. 2.Stability 0.5, similarity 0.75, speaker boost on
3. 3.Drop delivery tags into the text at section turns: [confident] at the open, [excited] at the turn, [serious] at the warning, [curious] at the close
4. 4.Use the with-timestamps endpoint so every word comes back with a start and end time. That timing drives every caption later

Monotone is a setting, not a limitation.

HeyGen builds a photo avatar from a single still and lip-syncs it to your audio. The whole result lives or dies on which frame you pick.

1. 1.Shoot a 30-second selfie clip talking to camera, or use one you already have
2. 2.Step through it frame by frame and find the one shot where your eyes are on the lens, not the screen
3. 3.Upload that frame as a photo avatar
4. 4.Generate the video from your 11 Labs audio: 9:16, 1080p

Second snag: the newer photo-avatar endpoint cost 177 credits for one clip. The legacy talking-photo path did the same job for 42. Same output.

HyperFrames renders video from an HTML file. Every element of the reference format becomes a component you build once: the hook card, the YouTube Studio mockup with your own thumbnails, the pill captions, the phone player, the watch page with a view counter.

1. 1.Captions are timed from the 11 Labs word list, so a new script re-times the whole video automatically
2. 2.The avatar video is the base layer, cropped differently per section
3. 3.Screen shares are HTML mockups of your own channel: real avatar, real titles, real view counts
4. 4.Run the check gate (layout, contrast, motion) before you render

npx hyperframes init my-short --example blank --resolution portrait
npx hyperframes check
npx hyperframes render --quality high --output out/short.mp4

A clean voice chain, a soft music bed carved under the voice, and a loudness target copied from the reference.

1. 1.Voice chain: high-pass at 90 Hz, mud cut at 250 Hz, gentle compressor, presence lift at 3 kHz, limiter at minus 1 dB
2. 2.Music: a low-volume bed whose energy sits under 80 Hz with sparse mids, so it never fights the voice
3. 3.Carve the bed so it dips in the bands the voice uses and comes back between phrases
4. 4.Measure the reference in LUFS and normalize to it. The reference reel measured minus 14.2, so that was the target
5. 5.Finish with a 1.2x speed pass, pitch preserved

Zernio gives you one API key, one profile, and OAuth links to connect YouTube and Instagram. One script uploads the MP4 and posts to both.

1. 1.Connect each platform through the API-generated link, not the dashboard button
2. 2.YouTube detects Shorts on its own: vertical and under 3 minutes
3. 3.Title under 100 characters, keyword first. Category set to Education, because the default is People & Blogs
4. 4.First three hashtags in the description show above the title
5. 5.Set Related Video on the Short to your long-form or your VSL. It is the free traffic button most people never flip

On Instagram, one comment-to-DM rule on the reel. Anyone who comments your keyword gets the promised resource by DM automatically, plus a public reply so the thread shows it works.

1. 1.Keyword match on the whole word, typo tolerant, also fires if they DM the keyword instead of commenting
2. 2.Two DM variants rotate so repeat sends do not look templated
3. 3.YouTube has no DMs, so pin a comment with the keyword instead

The keyword has to deliver something you already have. A promise you have to scramble for is not a lead magnet.