Use cases
Worked examples on Sume's catalog Formats and endpoints: product video and photography, try-on, app and product ads, logo animation, and more.
Start with Ready-made Formats for product video: the Sume Format catalog
Gameplay Shorts after YouTube's originality update: what pages say
YouTube's reused-content pages don't name gameplay. What they say about your own footage plus commentary, and how to cut highlight Shorts where you talk.
Griffin-Lite safety hold: what to prepare now for AI video calls
Tavus is holding Griffin-Lite over deception risk. Three things to prepare before any realistic AI presenter ships: consent, disclosure and a recorded fallback.
Carousel music is blocked with video slides: use a frame still
Instagram says music cannot be added to carousels with videos. Pull a still from each video slide with Sume video frames to keep the music option.
Per-slide carousel captions: pair each image job with its caption
Instagram added per-slide carousel captions. Keep each slide, its caption and its Sume image job in one manifest so a retry never mismatches. Script included.
Instagram posts in Google: draft the caption from the transcript
Public pro-account posts can appear in Google. Write the caption from what the video says; Sume video inspect returns a transcript at $0.01 per audio minute.
Instagram Replace Audio on a published post, or re-render with Sume
Instagram added audio replacement on published feed posts. Sume can render a new soundtrack but cannot feed it into that flow. Which route keeps what.
Reel safe zones: set compose margin_ratio to keep text clear
Instagram Edits shows safe zones in the editor. For a plate added in code, Sume compose overlay margin_ratio (0 to 0.45) sets the distance from the edge.
Lipstick shade preview with image edit: what a render cannot promise
A shade preview from gpt-image-2.5 on Sume is a picture, not a colour match. How to build it with a mask, and what to tell shoppers.
Ad from one product photo: a 4:5 still, a 9:16 clip and a 6 s cutdown
Three Sume calls turn one packshot into a 4:5 feed image, a 9:16 UGC-style clip and a short cutdown. Settings, order and what to check at each step.
Meme and trend Shorts: when many creators upload the same thing
YouTube lists content uploaded many times by other creators as reused content. How to join a trend with something of your own, using trim and captions.
Move an object in a photo with AI: erase, then place with two masks
No Sume image model has a move operation. Move an object by erasing it with one mask and placing it with a second, using GPT Image 2.5 and Python.
Movie and TV clip Shorts: what YouTube's help calls original
Unedited movie or TV clips are ineligible on Shorts. What YouTube's pages allow instead, and how to cut trailer-style Shorts from your own film or series.
Music for an avatar video: pass the preview still as image_url
Approve an avatar preview, then send its preview_image_url to the Music Router as image_url so the track matches the frame. Steps, limits and the $0.125 cost.
One garment photo to five shop assets: a small clothing shop workflow
A small shop with one photo per garment can make a cut-out, on-model still, 4:5 post, 9:16 clip and try-on with Sume calls. Order and checks.
Nine portraits in one AI image call: 3x3 grid prompt and slicer
Generate a 3x3 grid in one request and slice it into nine tiles. The pixel math at the 8.29 MP cap, what you give up in resolution, and when nine calls win.
Product review Shorts per SKU: vary the focus, not just the name
YouTube allows a review channel when each video has a distinct focus and rejects minimal variation. How to plan a batch of per-product Shorts for that test.
Recast or reshoot: break-even for ad variants, with Python
Recast costs source seconds times $0.375 at 768p on Sume. Enter your own reshoot cost per variant and a short Python function returns where Recast wins.
Replace an actor in a video with AI: what Recast keeps and swaps
To replace an actor in a video with AI on Sume, send the clip and one photo per person to h3-max-recast. Motion, camera, cuts and sound stay; people change.
Same intro and outro on every Short: YouTube lists it as allowed
YouTube's inauthentic-content page allows one intro and outro if the bulk differs. Build a fixed intro, changing body and outro with Timeline.
Saved favorites to a lookbook sheet: one gpt-image-2.5 call
Turn up to 16 saved product images into one lookbook sheet with openai/gpt-image-2.5 on Sume. Reference numbering, layout wording and what to check.
Screen-recording tutorial Shorts: what YouTube expects you to add
A screen recording with no voice is thin under YouTube's originality rules. How to cut a tutorial Short from your own recording with trim and caption cues.
Share another creator's video on YouTube without re-uploading it
YouTube points creators who want to show another creator's video to the Posts tab. How the share works, its limits, and an own-footage Short for your take.
A speaker withdrew voice consent: finding every narration that used it
Revocation is a clean-up job. Build a manifest of voice, model and job id for each narration, then regenerate with another voice and join audio on Sume.
Sports replay Shorts: YouTube allows them if you explain the moves
YouTube lists sports replays where you explain what the competitor did as allowed. How to build a vertical breakdown Short from your own match footage.
Stock footage and an AI voice: what makes a Short yours on YouTube
YouTube lists image slideshows and generic AI templates with minimal narrative as not allowed. What your script and edit must carry with stock clips.
Swap a person in a 4-second video: Recast needs 5 seconds
A 4-second clip is below h3-max-recast's 5-second minimum. Genjutsu takes 4 to 30 s, Gemini Omni edit 3 to 10 s, and Avatar Face Swap about 4 to 15 s.
Swap the presenter in a product demo for each market with Recast
Plan one source demo and several Recast jobs, one per presenter photo: cost for five variants, idempotency keys, and the voice caveat, using documented limits.
Sync music to avatar scenes: hook, demo, CTA timestamps in the prompt
Match a Lyria bed to a 12-second multi-scene avatar video by writing [0:00-0:03] section markers into the Music Router prompt. A worked example and cost.
Ten reference images for one product still: what goes in each slot
FLUX 3 Image takes up to ten references, GPT Image 2.5 on Sume takes sixteen, Ideogram 4.5 five. A slot plan and a script that trims it to each model's cap.
Text-only Shorts on YouTube: when scrolling text counts as value
YouTube lists scrolling text with minimal narrative as not allowed. How to write a text Short with real commentary and time its cues with Sume captions.
TikTok Ad Network carousel: 50 images max, pull stills from a video
TikTok Ad Network carousel and image ads take up to 50 images at 1200x628, 640x640 or 720x1280. Pull 24 stills per call with Sume video frames.
TikTok Ad Network video: up to 10 minutes, 500 MB, 720x720 allowed
TikTok Ad Network (Pangle) video ads accept 1280x720, 720x1280 or 720x720, up to 10 minutes and 500 MB. Cut each size with Sume video trim.
TikTok catalog add-ons: 130px card, 546px tile, keep the bottom clear
TikTok's interactive add-ons take a 130px or 546px band at the bottom of the video. Check a clip with Sume video frames; lift captions with anchor_ratio.
TikTok Display Card image: upload 1920x1080, shown at 750x421
TikTok Display Card: JPG or PNG, 1920x1080 recommended, shown at 750x421, popup time set by TikTok. What to make and how to pick a contrasting color.
TikTok playable ads: 5 MB zip, no HTTP requests, shrink the video
A TikTok playable ad is a zip under 5 MB with index.html and config.json, and no outside requests. Sume builds no playable, but can shrink the clip inside one.
TikTok Symphony Agent makes 3 variations at a time: run three cuts
Symphony Agent outputs up to 3 video variations at once for A/B tests. How to produce three cuts from one master with Sume trim, with a separate key for each.
Trim before you recast: pay for the seconds the swap needs
Recast bills the whole source length. Cut a 30-second clip to the 12 seconds that matter with POST /v1/video-trim and the 768p price drops from $11.25 to $4.50.
Vision Pro app preview 3840x2160: Sume trim stops at 2160 per edge
Apple asks for 3840x2160 for Apple Vision Pro previews, but Sume video trim caps output edges at 2160. What fits for iPad and Mac, and what does not.
Voice clone consent record: what to keep before you upload audio
Microsoft says MAI-Voice-2.1 matches a voice from a few seconds of audio. An eight-field consent record, a checker that blocks gaps, and what Sume's docs cover.
Weekly AI release-notes video: 52 x 30 seconds, annual cost on Sume
A weekly 30-second AI presenter release-notes video for a year is 1,560 seconds: $382.20 on Plus, $287.04 on Standard, with a $6.50 music bed and $0.95 avatar.
YouTube daily upload limit for Shorts: no published number, pace it
YouTube's help says a daily upload limit exists across desktop, mobile and API but names no number. How to pace a Shorts series without guessing.
Ineligible Shorts views: the three kinds YouTube says do not count
YouTube's Shorts page names non-original, bot and advertiser-unfriendly views as ineligible. What each means and which of them you control.
Short with one licensed track: half the revenue covers the music
YouTube's Shorts page says one track means half the engaged-view revenue covers licensing. What it says, what it omits, and a no-library-track option.
Add fireworks or confetti to a video with Gemini Omni Flash edit
Try adding fireworks or confetti to a clip with a Gemini Omni Flash 1.1 edit on Sume: prompt wording, light matching, request and what to inspect afterward.
AI clip to a vertical Short, in order: captions go last
The call order that avoids rework when turning an AI clip into a YouTube Short on Sume: generate, probe, trim, dim, music bed, render, then burn captions last.
AI time-lapse video prompt for Gemini Omni Flash on Sume
Write a time-lapse prompt for Gemini Omni Flash 1.1: Google's guide on lighting and camera, a locked-off frame, 10-second limit, and a request to test at 360p.
AliExpress product image size: 1000px square, then what to ask Sume
AliExpress's Image Size Guide says at least 1000x1000 px, 1:1 (4:5 for apparel), under 2 MB, white background. The matching Sume image request.
Best Buy product shot guide: 1500x1500 and what to ask Sume
Best Buy's Digital Assets Product Shot Guide: 300 DPI, 1500x1500 minimum, white background, UPC file names. The matching Sume Image API request.