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
4:3 tablet video to 9:16: the 0.4219-width crop in video filter
Crop a 4:3 clip to 9:16 with video-filter: width 0.4219, x 0.2891, full height, then scale to 1080x1920. Numbers, rounding, and the free check.
Join TTS lines into one track: match output_format before concat
Sume timeline audio concat joins 1 to 20 parts but needs one channel layout. Set the same TTS output_format on every line, then concat once.
9:16 Short to a 1:1 feed clip: crop height 0.5625 with a y offset
Cut a square 1080x1080 clip out of a 1080x1920 vertical video with video-filter: full width, height 0.5625, y 0.1 for headroom. Numbers and request.
Seamless loop with Wan 3.0: the same image as first and last frame
Send one image as both start and end frame to ask Wan 3.0 for a loop on Sume, keep it 4 to 6 seconds, and check the join with stills before it repeats.
TikTok video limit: guides say 10 or 60 minutes, so cut to the lower
Third-party guides give TikTok 10 minutes in-app and up to 60 for upload. Probe the file with video inspect, then cut to the lower figure with video trim.
Tighten a stitched Omni cut: source_in and duration per timeline slot
Cut the first half-second off each Omni clip without a trim job: source_in and a shorter duration per Sume timeline slot, with the starts and spine math.
Tall phone screen recording to 9:16: crop height 0.8215 and a y offset
A 1170x2532 screen recording is taller than 9:16. Keep the full width, crop height 0.8215, and choose y to drop the status bar. Request and numbers.
720p source to 9:16: a 404x720 strip, upscaled 2.67x to 1080x1920
Cropping 1280x720 to 9:16 leaves a 404x720 strip. Scaling it to 1080x1920 enlarges it 2.67x. What that does to sharpness, and the alternatives.
End a vertical Short on a fade: Timeline fade_out_seconds 0.5
Timeline output.fade_in_seconds and fade_out_seconds run 0 to 5 s, and their sum cannot exceed the output length. Request, limits, and error code.
Hailuo 3.0 clips in a Sume timeline: set output.fps or omit it
Cutting 24 fps AI clips into a timeline at the wrong frame rate causes judder. Sume's Timeline output.fps accepts 24, 25, 30 or 60, or matches your sources.
Animate an infographic with Wan 3.0: first frame or reference image?
Use the infographic as a first frame to keep the layout, or as a reference to get new motion. How Sume decides, and which to pick for charts, icons and text.
Did my Wan 3.0 clip actually cut between shots? Check with stills
Wan 3.0 advertises auto scene splitting, so verify it: pull stills at named seconds with Sume video inspect, or read shot cuts with reference ingest.
Add a "save this" end card to a Short with caption cues
After the July 2026 Save button, burn a short on-screen ask into the last 3 seconds with video-captions cues: start, end, text. No speech needed.
TikTok safe zone: no single number, so test a caption anchor ratio
TikTok has no safe-zone figure I could read from its own help page. Place burned-in captions with design.placement.anchor_ratio, then check a still.
4K 3840x2160 to 9:16: a 1214x2160 strip, then 1080x1920
A 4K UHD clip cropped to 9:16 keeps a 1214x2160 strip, more pixels than 1080x1920 needs. The fractions, the scale-down, and the 300-second limit.
Fade 0.25 s between two vertical clips: Timeline transition limits
Timeline transitions run up to 1 second and half of the shorter neighbor, never on the first slot, and no more than 8 chained. Request and error codes.
Lyria music prompt rejected by policy: what to change, what to keep
After a policy rejection on a Sume music job, edit only the flagged content, keep the seven-axis brief, and retry in budget. A generic bed is the wrong fix.
Check on-screen text in a Wan 3.0 clip: video frames at exact seconds
Wan 3.0 claims text rendering in 12 languages. Before you ship, pull PNG stills at the seconds the text appears with Sume video frames and read them yourself.
Fix the text in a reference image before it goes into an H3 clip
Spend one cheap Ideogram 4.5 edit on a reference image's text before you spend on a MiniMax H3 clip that copies it. The order on Sume, with a request.
MiniMax H3 takes 9 image references: a slot plan for one product clip
Hailuo 3.0 (MiniMax H3) accepts up to 9 reference images per clip. A 9-slot plan for a product ad, and the request to send it on Sume.
Replace the music in a vertical Short: trim audio drop, then a bed
Remove a clip's audio with video-trim audio drop, then rebuild the track in Timeline with a silent spine and a soundtrack bed at a set gain_db.
Match voice emotion and music mood: one mood word, two Sume fields
Set generation_config.emotion on TTS and the emotion axis in the Music prompt from the same mood word, so a short's voice and bed do not argue.
3:4 portrait clip to 9:16: crop 0.75 of the width, scale to 1080x1920
A 3:4 portrait clip is already tall. Crop 0.75 of its width (x 0.125), keep full height, scale to 1080x1920. Numbers and the video-filter request.
Draft a Short at 720x1280 in Timeline, then render 1080x1920
Set Timeline output width 720 and height 1280 to review a vertical cut, then render the final at the 1080x1920 default. Same per-minute rate, checked by plan.
Remove the card behind burned-in captions: colors.card null
Set design.colors.card to null on a Sume caption job to show no card. It works on styles that support design, not on punch or tiktok-green.
Video captions take a public HTTPS URL; other media tools do not
Which URL each Sume video tool accepts: captions fetch a public HTTPS clip, while trim, filter, frames, inspect and timeline need your media.sume.com file.
Captions with no style set: Korean gets black-outline, no gold tint
If you send no style, Sume picks one from the caption text, and the spoken word gets no gold tint. Set design.colors.active to add one.
Continue a Wan 3.0 clip: trim the tail, send it as a reference video
Wan 3.0 takes reference videos on Sume (up to 5, 15 s total). To continue a scene, cut the last 3 seconds with video trim and use them as the next reference.
Caption a silent product video: caption_no_speech and cues
A silent clip fails speech-to-captions with caption_no_speech. Send cues with text, start and end to burn authored overlay text on /v1/video-captions instead.
Turn a music composition plan into a Sume time-range prompt (Python)
ElevenLabs music_v2_5 plans allow 6,132 characters in up to 30 lines. Sume's Music prompt takes 5000 characters. A Python converter for time ranges.
Lyria 3.5 output: MP3 or WAV at 44.1 kHz, and what Sume returns
Google lists MP3 by default or WAV, 44.1 kHz stereo, with a SynthID watermark. Sume's music docs say the artifact is typically audio/mpeg. Check it in code.
Lyria 3.5 prompts: [Verse] tags or [0:00-0:30] time ranges?
Google's Lyria 3.5 docs show [Verse], [Chorus] and [Bridge] tags. Sume's music docs show time-range markers. What each says, and how to test both on one brief.
Stability AI's Series B and label backers: what it means for audio
Stability released Stable Audio 3.0 on 5/20/26 and raised a Series B on 8/25/26 with EA, Sony, UMG and WMG. A dated timeline and what to verify for video work.
Cut a clip into 30-second chunks for 30-second post-processing caps
Runway's SDR-to-HDR model and the Magnific upscaler take 30 seconds at most. Cut any clip into 30-second pieces with Sume video trim at $0.02 per cut.
Seedance 2.5 secondary edit vs trim-and-regenerate on Sume
BytePlus describes timestamp-level edits to Seedance 2.5 clips. Sume has no such edit field, so here is the trim-and-regenerate route and its limits.
Transparent PNG: generate it or cut it out?
Sume lists a background option on ChatGPT Image 2.5: auto, transparent or opaque. Generate a transparent PNG directly, or cut out an existing image.
WhatsApp Status 16 MB and 30 s: trim, then check size
A third-party guide lists WhatsApp Status video at 16 MB and 30 seconds. Cut to 30 seconds with Sume video trim, then measure the file size.
FTC insider reviews: burn the disclosure into the clip
FTC rule section 465.5 requires a clear and conspicuous disclosure of insider reviews. Burn a disclosure line into the clip with Sume's video captions cues.
Caption a silent AI video: fixing caption_no_speech
A silent clip fails POST /v1/video-captions with caption_no_speech. Send cues with text, start and end to burn authored captions without speech-to-text.
Temu detail video up to 180 s: stitch 30 s clips with Timeline
Temu detail video allows up to 180 s and 300 MB at 720p or higher. Sume clips top out at 30 s, so six clips stitched with Timeline cover the full length.
LinkedIn video thumbnail under 2 MB: pull the frame with Sume
LinkedIn video ad thumbnails are JPG or PNG up to 2 MB. Pull a still from your clip with Sume's video frames endpoint and check its size before upload.
MAI-Voice-2.1 Flash narration for a video cut
Narration from any voice model can be the audio spine of a Sume Timeline cut. Import the file first, then render at $0.10 per output minute.
SeedVR2 10x video upscale vs Sume's upscale tool
SeedVR2 is listed with 1x to 10x video upscaling. Sume exposes a video_upscale_create tool, but its docs name no scale factors, so check the live schema first.
Pull a voice track from AI video, then join takes gaplessly
Extract a wav or mp3 from a Sume-hosted video with POST /v1/audio-detach, then join up to 20 audio parts with Timeline audio. Fields, caps, $0.01 rate.
Edits carousels: four matching images per call
Instagram Edits now supports carousels, per secondary reports. Sume image models with an n range of 1-4 return up to four matching images per request.
Remove a video background: Firefly vs Sume options
Adobe Firefly can now remove a video background across a whole clip. The Sume docs list a cutout tool for stills, not video, so plan a clean backdrop instead.
Indiegogo video 2-4 min: assemble 30 s clips in Timeline
An Indiegogo main video runs 2-4 minutes. Generated clips top out at 30 s on some Sume models, so assemble them with Timeline 1.0 at $0.10 per output minute.
Douyin ad first frame: black at most 60%, check with stills
Douyin in-feed ads need a first frame that is at most 60% black. Pull stills from your render with Sume video inspect and measure the first frame before upload.