Media tools
Sume media tools by API: trim, filter, captions, frames, inspect, speech-to-text, upscaling, background removal, transitions, and Timeline assembly.
Start with Video editing API limits: Sume media tool caps in one table
Change a video's background with a prompt: Gemini Omni on Sume
Google pitches Gemini Omni background changes by prompt. How to swap the background of your own clip with Sume's video edit mode: request and limits.
Cut out a subject from a video frame: video frames, then RMBG
Pull a lossless PNG frame with Sume video frames (unbilled), then run RMBG 1.0 on it for a transparent PNG. Sume does not remove backgrounds from a whole video.
Cut out then upscale, or upscale then cut out? Order for PNG alpha
Upscale first, then remove the background last so nothing can drop alpha. Sume RMBG returns PNG with alpha; upscale takes output_format png, jpg or webp.
Flow Insert: add an object to a video, and the Sume API equivalent
Flow's Insert adds objects to a generated scene. What Google's launch post says, and how to add a prop to an existing clip with Gemini Omni video edit on Sume.
Does your Instagram Reel loop? Compare first and last frames
A Reel loops cleanly when its last frame leads into its first. Pull both stills with Sume video_frames, then re-cut the end with video trim.
Reel shared to feed: the cover is cropped to the middle square
With share_to_feed, Instagram uses the middle 1:1 of your cover for the feed post. Preview it with Sume video filter crop fractions and keep text inside.
Instagram Reels captions too small or cut off: tune safe width
Resize, narrow and move burned-in captions per request with Sume video captions design: font_size_ratio, safe_width_ratio and anchor_ratio, with their ranges.
Steam trailer thumbnail: a 1920x1080 frame from the video itself
Steam wants the trailer poster as a 1920x1080 JPG or PNG taken from the video. Pull it with Sume video frames at max_edge 1920, plus the trailer specs to check.
Telegram video emoji 100x100: Sume's 256 px floor and a square crop
Telegram video emoji must be exactly 100x100. Sume video trim outputs 256 px at minimum, so crop a square and cut in Sume, then downscale and encode outside.
Telegram video sticker: 512 px, 3 s, 256 KB, WEBM VP9 with Sume
Telegram video stickers need a 512 px side, 3 s, 30 fps, 256 KB, no audio, WEBM VP9. Sume trims and resizes but outputs MP4, so one step stays outside.
Tenor API closed to new clients: make short looping clips instead
Google's Tenor quickstart says it stopped taking new API clients in Jan 2026. Sume does not search GIF libraries, but it can trim your own MP4 clips.
Video upscale tops out at 30 s: trim the hero seconds first
Sume video upscale takes 1 to 30 input seconds and bills per input second. Cut the best few seconds with video trim first, then upscale only those.
YouTube Shorts 16:9 thumbnail swapped for 4:5: make a safe one
YouTube swaps a 16:9 custom thumbnail on a vertical video for an auto 4:5 on home feeds. Preview the 4:5 crop with Sume video filter and video frames.
YouTube Shorts max resolution is 1080p: skip the 4K upscale
YouTube Help says Shorts upload at a maximum of 1080p and up to 3 minutes. Upscale a 720x1280 clip by 1.5 with Sume to reach 1080x1920, and stop there.
YouTube thumbnails.set costs about 50 quota units: plan a batch
Google lists about 50 quota units per thumbnails.set call, a 50MB cap, JPEG or PNG. Sume video frames is unbilled, so spend calls on one chosen still.
Make a video black and white with the API: monochrome filter
To make a video black and white in Sume, send filtergraph monochrome to video-filter. It is on the allowlist and returns a new MP4. Source untouched.
Bria increase_resolution (2x or 4x, 8192 px) vs Sume image upscale
Bria's increase_resolution takes desired_increase of 2 or 4, up to 8,192x8,192 px. Sume Image Upscale 1.0 takes any upscale_factor from 1 to 4, default 2.
Bria remove background preserve_alpha vs Sume RMBG on cutouts
Bria's remove_background has preserve_alpha and moderation flags. Sume RMBG 1.0 has neither; it flattens already-transparent inputs onto gray before cutting.
Burn Japanese, Chinese or Arabic captions: what Sume documents
Sume's caption docs list Latin and Hangul faces only. For other scripts, test one clip first, or ship a sidecar file where the platform takes one.
Caption a dubbed video: new job, not source_caption_id
source_caption_id reuses the first video's word timings, so a dub needs its own caption job on the dubbed file. How Sume handles it, and the cost.
Check Seedance or Kling dialogue with a video-inspect transcript
Did your Seedance or Kling clip say your line? Run video-inspect with transcribe true on the Sume-hosted clip and compare the words. $0.01 a minute.
Crop 16:9 to 4:5 with video-filter: width 0.45 and an x offset
To crop 16:9 to 4:5 in Sume video-filter, use width 0.45, height 1 and an x between 0 and 0.55. Here is the math, the x choice and the output scale.
Cut a Seedance 2.5 clip shorter: video-trim or regenerate?
To shorten a Seedance 2.5 clip, cut it with video-trim at $0.02 a job. Regenerating re-rolls the whole video and cannot go under the model's 4-second minimum.
Fix a wrong display aspect ratio: setdar in Sume video-filter
If a video plays stretched, send setdar to Sume video-filter: setdar is on the allowlist, setsar too. Use check first, then encode a new MP4.
Flip a video upside down with the API: vflip in video-filter
To flip a video upside down in Sume, send filtergraph vflip to video-filter; hflip mirrors it. vflip, hflip and transpose are all on the allowlist.
Sume image upscale output_format webp: accepted, but you get a PNG
Sume Image Upscale 1.0 lists png, jpg and webp for output_format, but only JPEG or PNG come out: webp falls back to PNG. Check content_type.
Localized caption batch: eight QA checks before you ship
Eight checks for a batch of burned caption videos in several languages: script coverage, language hint, timing, key per job, and cost. Sume job behavior.
LTX-2.5 Alpha Gen: video mattes without a green screen
LTX-2.5 Alpha Gen turns RGB clips of up to 145 frames into grayscale mattes. What it needs, where it stops, and what Sume's RMBG route does instead.
LTX-2.5 Restore LoRA for archive footage vs Sume video upscale
LTX-2.5 Restore cleans and colorizes damaged archive clips with tiled 8-step runs. Sume Video Upscale enlarges by 1.1 to 4 times and does not colorize.
LTX-2.5 SDR-to-HDR LoRA: ACEScg EXR and HLG, not an MP4 API
LTX-2.5 SDR-To-HDR outputs scene-linear ACEScg EXR frames plus an HLG master from an SDR clip. Sume's video routes return MP4. Where each fits.
Luma Photon image reframe: 10 MB source, 2432x1024 max vs Sume
Luma's Reframe API extends an image to a new aspect ratio with photon-1 (10 MB in, 2432x1024 out at most). Sume has no reframe call; here is the closest path.
Pika SFX: 1-20 second clips, negative prompts and seed vs Sume
Pika SFX makes 1 to 20 second effects with negative prompts, seed and guidance. Sume lists no sound-effects model, so here is what to use for each need.
Skim a 30-second Seedance clip with video-inspect seek fast
To skim a long Seedance 2.5 clip, call video-inspect with frames fps and seek fast. Stills snap to keyframes, a few seconds early at most. Stills are free.
Starlight Precise 2.6 sharpness levels vs Sume video upscale tiers
Topaz Starlight Precise 2.6 adds a three-level sharpness slider. Sume Video Upscale 1.0 has no sharpness setting: it has fast, standard and pro tiers.
Veo upscaling on Vertex works on any video: Sume's 1.1x to 4x job
Google's Veo upscaling on Vertex AI is a private preview that takes any source video up to 4K. Sume's upscale job takes a public HTTPS video, 1.1x to 4x.
Video trim end past the clip: trim_clamped_to_source vs start error
A Sume trim whose end runs past the clip clamps and warns trim_clamped_to_source. A start past the clip fails with trim_start_past_source.
Sume video upscale: scale_ratio overrides upscale_factor
Both fields set the video upscale multiplier, 1.1 to 4. When you send both, scale_ratio wins. Omit both and you get 2x. Here is how to send it and check.
YouTube Create Omni lighting options: reuse them as prompt words
YouTube Create offers four lighting options for Gemini Omni clips and caps prompts at 900 characters. How to write the same looks into a Sume 9:16 request.
YouTube 'Make me move' and Photo to video: limits and a Sume route
YouTube's Photo to video and Make me move turn a photo into a Short clip of up to 10 seconds. Where they're missing, and how to animate a photo with Sume.
YouTube Shorts Align tool: continue a scene from the last frame
The Shorts Align tool overlays the last frame so the next clip lines up. Do the same with Sume: pull the last still, then use it as a first frame.
AI music brief template: seven axes for a Lyria 3.5 prompt
A seven-part music brief for Lyria 3.5 on Sume: emotion, genre, BPM, key, instruments, arc and era, with a filled example and the fields the API rejects.
Check a restyled video kept the camera move: pull matching frames
A restyle should keep motion and camera. Extract frames at the same times from source and output with Sume's unbilled video frames endpoint and compare.
Edit part of a product ad image with a mask: ChatGPT Image 2.5
To change a product ad's background without touching the product, send the photo and a mask_url to ChatGPT Image 2.5 on Sume. Fields and limits.
Check every Meta ad crop for cut-off text with video frames
With ad-set placement exclusions removed, each crop can run anywhere. Pull stills from every cut with Sume video frames and check the price and headline.
Add captions to a MiniMax H3 clip: video-captions, silent clips
A minimax-h3 clip carries native stereo audio, so video-captions can transcribe it. Silent clips fail with caption_no_speech; pass cues instead.
Pull the audio out of a MiniMax H3 clip with Sume audio-detach
A minimax-h3 clip has native stereo audio. audio-detach returns it as a wav or mp3 artifact without touching the video. Fields, defaults and caps.
Read on-screen text in video: stills at 768 px, or raise max_edge
Video inspect returns 8 stills at 768 px by default. For small on-screen text, raise max_edge up to 2160 or ask for PNG. Settings, limits and cost.
Static ad aspect ratios an image API can return: Sume's list
Which aspect ratios Sume's Image API accepts for static ads, why 4:5 is not 4:3, and why you read the model's own list before pinning one.