Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Kling motion control with mode webhook: a Python verifier for rotation
Submit a Kling 3.0 motion control job with mode webhook and verify the sume-v1 signature in Python, including the two-entry header during a secret rotation.
LinkedIn webinar teaser, 15 seconds: Seedance 2.5 vs Wan 3.0 cost
A 15-second webinar teaser costs about $4.03 on Sume's seedance-2.5 at 480p and $1.88 on wan-3.0 at 720p. Here is what that gap buys you.
List the Formats by Sume that return video, in Python
Call GET /v1/formats, page with next_cursor, and filter on the sume handle and io.output_kind to find ready-made video Formats. Standard library only.
Luma Ray3.2 face tracking vs Kling motion control: pick by input
Luma Ray3.2 tracks up to eight faces frame by frame; Kling motion control moves a still with a reference clip. Which to use, and what Sume runs today.
lyria-3-pro-preview vs Sume lyria-3-pro: which model string?
Google lists lyria-3-pro-preview at $0.08 per song. On Sume the Music Router id is lyria-3-pro, and a Google string fails with model_not_found.
Macro liquid pour video prompt for Gemini Omni and its price
A prompt recipe for a macro pour shot (oil, serum, honey) with Gemini Omni Flash on Sume: camera words, sound line, 5-second cost at 360p, 720p and 1080p.
MAI-Transcribe-2-Streaming partials vs Sume STT terminal webhook
MAI-Transcribe-2-Streaming sends intermediate and final results as audio streams in. Sume STT sends one signed terminal callback. How to design for each.
Mobile pet groomer reel from one photo: Seedance 2.5 image-to-video
Turn one photo of a groomed dog into a vertical reel clip with Seedance 2.5 on Sume: first-frame input, 4 to 30 seconds, native audio, async job.
Name your avatars: a handle scheme that fits 2 to 30 characters
Sume avatar handles allow letters, digits, periods and underscores, 2 to 30 characters. Build a team naming scheme that passes validation and stays readable.
Nano Banana 2.1 tracks 4 characters, 10 objects: Sume's cap
Google says Nano Banana 2.1 tracks 4 characters and 10 objects across up to 14 references. Sume's Image API caps input_references at 10, or 16 on GPT.
Gemini Omni 'no dialogue' prompt: steer the audio on Sume
Omni clips always carry sound. Google suggests 'no dialogue' and 'no extra sound effects' in the prompt. How to use that on Sume and what a test costs.
Gemini Omni sound effects prompt: name each sound in 8 seconds
How to prompt footsteps, a door and breaking glass in a Gemini Omni clip: Google says describe audio explicitly. Sume request, timing and cost per take.
Let users pick an avatar in your app with GET /v1/avatar-1.0/avatars
Build an avatar picker on the Sume Avatar 1.0 list route: read ready avatars server-side, cache them, and pass the chosen handle to talking-video.
Pick an image model by job: product photo, poster text, edit
Five new image releases in a month. A job-first table for Sume: GPT Image 2.5, Ideogram 4.5, Nano Banana 2.1, FLUX.2 and Qwen, with the limits that decide.
POST /v1/images returns 200 or 202: one Python handler for Sume
Sume's image route blocks up to 30 seconds and returns 200 with images, or 202 with a job envelope. One handler that covers both and never double-submits.
Sume job webhook receiver in plain Python, no framework
A 30-line stdlib http.server receiver for Sume job webhooks: raw-body HMAC, 300-second window, rotation-safe header, refuses an empty secret, answers 204.
Qwen Image 2.1 Pro: what exists, and which Qwen ids Sume lists
People search for Qwen Image 2.1 Pro. Its model card lists no Pro variant; Sume's Image API lists qwen/qwen-image and qwen/qwen-image-max.
Read the motion clip length with video inspect before Kling duration
Kling motion control on Sume reserves money from the duration_seconds you declare. Probe the reference clip with video inspect first, so the number is measured.
Restaurant dish reveal video prompt for Gemini Omni, 6 seconds
A six-second plating and steam reveal for a menu item with Gemini Omni Flash on Sume: camera move, audio line, first-frame trick and price at 720p and 1080p.
Retexture a product photo with FLUX.2 pro: prompt and call
Change leather to canvas or oak to walnut while the shape stays put. A FLUX.2 pro reference edit on Sume: one photo in, a preserve clause, aspect_ratio auto.
Runware deliveryMethod async vs Sume mode async: map the fields
Runware returns a taskUUID for async video, with getResponse polling or a webhookURL. Sume's equivalents: a job id, /v1/jobs/:id/status and mode webhook.
Scheduled or Format API: does the clock or your user start the run?
A Sume Scheduled run fires on a cron; a Format run fires when your backend calls it. Same agent, same receipt, new trigger. How to choose without dupes.
Shorts-only YouTube playlists can't be reordered: render in order
YouTube's playlist help says reordering is unavailable for Shorts-only playlists. Fix episode order before upload and render in order with fixed keys.
Turn a shot list into Gemini Omni requests: Python on Sume
Submit every line of a shot list as its own Omni job with stable Idempotency-Keys, then poll them in one pass. Runnable Python and the cost of a 3-shot list.
Skippable in-stream bills at 30 seconds: cut a 30-second version
Google bills skippable in-stream per view at 30 seconds watched, or full length if shorter. Cut a 30-second version of a long master with video-trim.
Slow-motion water splash prompt for Gemini Omni: what 'slow' does
Prompt a slow-motion splash in Gemini Omni: how to word the speed, why a 5-second clip is enough, and what 360p, 720p and 1080p cost on Sume.
Split a 9-minute recording into three Reels under 3 minutes
Instagram's Reels page says Reels over 3 minutes won't be recommended to new audiences. Find sentence ends with video-inspect STT, then cut with video-trim.
No progress percent in Sume job events: what to log and show instead
Sume's job events are a public timeline of eight named events, not a progress feed. What each means, what to log, and what to show a user while a video renders.
Sume music job in Python: 30-second wait, then poll, one paid job
Submit a Lyria track to Sume Music Router with a 30-second sync wait, fall back to polling on timeout, and reuse one Idempotency-Key. Runnable Python.
Sume TTS without a language field: English default, Hangul fallback
What Sume TTS 1.0 does when you leave out language, why a Hangul-only script is the one fallback, and why you should set language yourself on every job.
Sume TTS sentence slices need wav or raw output, not mp3
Sume TTS returns per-sentence audio_url slices only for wav or raw output. With mp3 you get timings but no slices. The request that gets clips.
Sume video mode: async, sync, subscribe or webhook? Cost is the same
Mode only decides how you learn the outcome: all four create the same job at the same price. A decision table for web apps, workers and batch pipelines.
Sume webhook.test has no job_id: keep it out of your job table
The dashboard's Send test posts a signed webhook.test with no job_id. Route on event first, dedupe on job_id second, and no phantom job row appears.
Three-minute Short: songs cap at 90 seconds, so plan the music bed
YouTube says most songs run up to 90 seconds in a 3-minute Short, some 60 or 30. Build the 180-second timeline with a voice spine and a looping, ducked bed.
Three ways to run the Sume video agent from code
Format runs, Scheduled runs and Agent Completions all start the same Sume agent. Pick by how often your task changes, then read the receipt the same way.
TikTok non-Spark ad captions: no links, @ or hashtags, lint first
TikTok's auction page says non-Spark captions allow no links, @ symbols, or hashtags; Spark Ads take 4 caption lines. Lint the text in Python before upload.
White backgrounds hide TikTok's ad UI: dim the clip, check for free
TikTok's reservation page warns that white or transparent backgrounds can hide the nickname and captions. Dim with video-filter ops, check first, then re-probe.
TikTok reservation ads must include audio: probe, then add a bed
TikTok's reservation spec says every video creative must include audio. Probe has_audio with video-inspect, then give a silent clip a voice or music spine.
TikTok trending this week: a small shop's research call
Use Sume's trending-videos search with window this-week to see which TikTok videos in your niche are moving before you make one. Fields, price note, limits.
Video generation MCP connector: six questions before you connect
Before you let an agent spend money through any MCP connector, ask six questions. Here are Sume's hosted MCP answers on auth, scopes, idempotency and spend.
Voiceover plus a Lyria bed for a reel in one agent session over MCP
Ask an MCP client to run tts_create and music_create on Sume, then check the take with stt_create. Which tool does which job, and where the agent must wait.
Wedding venue teaser music: write the prompt with section markers
Score a 30-second venue teaser with Sume's Music Router: no duration field, so length goes in the prompt with timed section markers. Example prompt and request.
Hackathon app on the Sume Free plan: 1 seat, 6 accepted jobs
A weekend demo on Free can have one job processing and five queued. How to design the UI, the retries and the demo script around that, with the real limits.
Which Sume speech endpoint: TTS 1.0, TTS Router, STT or Music?
A decision table for Sume's four audio endpoints: TTS 1.0, TTS Router, STT 1.0 and Music Router. Inputs, limits, ids and what each does not do.
Will my upload become a YouTube Short? Check the probe first
YouTube classes square or vertical video up to 3 minutes as a Short; 16:9 stays long-form. Check width, height, rotation, and duration from the probe.
Wine bottle reveal clip: 6 seconds of Wan 3.0 from a label photo
Animate a bottle photo into a 6-second reveal with wan-3.0 on Sume: 38 cents at 480p, 75 cents at 720p, $1.50 at 1080p. Why the label needs checking.
YouTube Shorts max out at 1080p, so don't generate at 4K
YouTube's Shorts help page caps uploads at 1080p. Sume's Gemini Omni Flash 1.1 offers 360p to 4K in 9:16, so request resolution 1080p and skip the higher tier.
Cost of 10, 100 and 1,000 five-second AI clips on Sume
A five-second clip bills 3.75 to 37.5 cents per second on Sume, so $0.19 to $1.88. The budget table for 10, 100 and 1,000 clips on Wan 3.0, H3, H3 Max and Omni.