Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Validate clip length against supported_durations before a Sume submit
Durations differ by model: some start at 2 seconds, some stop at 10, some reach 30. Read supported_durations from the catalog and snap a request in Python.
Veo 3.1 price per second vs Sume's Video Router models
Google lists Veo 3.1 at $0.05 to $0.60 per second. Sume does not offer Veo 3.1 in its v1 catalog; here is the 8-second, 100-clip math for what it does.
A video agent run takes 15 to 30 minutes: design the waiting
Long-form video from an agent is minutes of work, not seconds. Email-me-when-ready, saved drafts and honest limits for a Sume Format run in your product.
AI voiceover cost for 100 scripts: Sume vs ElevenLabs v4
100 scripts of 800 characters cost $3.80 on Sume TTS, $6.40 on ElevenLabs v4 at list, or $1.76 under its discount that ends Oct 12.
waitForJob polls every 2 s: 600 reads in a 20-minute Sume video job
The SDK's waitForJob floor is a 2-second poll and a 20-minute timeout: at most 600 reads per job. The read budget by plan and why next_poll_after_seconds wins.
Wan 3.0 per-clip cost: Sume API vs Runway credits
Wan 3.0 is $0.0625 to $0.25 per second on Sume. Runway's page shows 5 to 20 credits per second at $0.01 per credit. Per-clip math inside.
Wedding save-the-date video: photo to end card with Gemini Omni
Make an 8-second save-the-date by pinning a couple photo as first frame and a designed date card as last frame in Gemini Omni on Sume. Request and cost.
Save-the-date video from two photos: Wan 3.0 first and last frame
A 5-second save-the-date that travels between two photos costs 63 cents at 720p on Sume's wan-3.0. How frame_images works and what to avoid.
Which TikTok ad placements take a 10-minute video? Only auction
TikTok's auction non-Spark page allows up to 10 minutes; reservation in-feed and TopView stop at 60 seconds. Pick the placement first, then cut with video-trim.
Write the music brief after the voiceover, using TTS timestamps
Lyria on Sume takes no duration field, only time ranges in the prompt. Generate the voiceover first, read its word timings, then write the section markers.
Sume yearly vs monthly plan price: the 10, 12.5 and 15 percent math
Yearly stickers are $432, $1,260 and $4,080 against $480, $1,440 and $4,800 monthly: 10, 12.5 and 15 percent off, or $36, $105 and $340 a month.
YouTube in-feed ads have no max length, but video-trim caps at 1,800 s
Google lists no max length for in-feed video ads, but Sume's video-trim takes a source of 1,800 seconds at most and outputs 900. Plan long cuts around that.
Airbnb house-rules video: an avatar in front of your living room
A short-term rental host can make a 9:16 house-rules video by placing a Sume avatar in front of a photo of the unit, using the talking-video photo scene input.
ASMR-style AI video prompt: Gemini Omni with sound, 10 seconds
Prompt Gemini Omni for ASMR-style clips with close hand movement and quiet foley. Sume request, why audio cannot be switched off, and what 10 seconds costs.
Five-lesson avatar onboarding series: one handle, safe retries
Plan an onboarding series of short Sume avatar videos: one handle, one script and idempotency key per lesson, a preview first, and captions on the final MP4.
C# HttpClient: submit a Seedance 2.5 job and poll until completed
A .NET top-level-statements script that posts a seedance-2.5 job to Sume, polls every 30 seconds, and stops on completed, failed or cancelled.
Claymation stop-motion look: a Gemini Omni prompt that holds
Prompt Gemini Omni for a claymation or stop-motion look: material words, a choppy-frame cue, and sound. Includes a 6-second Sume request and the cost per take.
Coffee roaster iced-coffee ad: GPT Image 2.5 still, then a clip
A coffee roaster can design one hero still with GPT Image 2.5 on Sume, check the bag and glass, then use it as the first frame of a short video. Two steps.
Cost per correct poster: Ideogram 4.5, GPT Image 2.5, Nano Banana 2.1
Price per call hides retries. Count images with every word right, divide the Sume usage.cost sum by that count, and compare three new models. Script included.
Sume default queue capacity: max(3, 5 x concurrency), checked per plan
Sume's default queue capacity is max(3, concurrency x 5). Checking the rule against Free, Pro, Startup and Scale, plus the org floor and the wave hint.
Dentist office welcome video: a 9:16 AI avatar clip from one script
A dental practice can make a vertical welcome video without a shoot: one avatar, one script of 4 to 60 seconds, burned-in captions, and a staff review step.
Dolly zoom (vertigo effect) prompt for Gemini Omni, tested cheap
Ask Gemini Omni for a dolly zoom by describing both motions, try it at 360p for 30 cents, and check the background stretch before paying for 1080p on Sume.
Edited script, same Idempotency-Key: 409 conflict on Sume TTS
Reusing an Idempotency-Key with a changed transcript returns 409 idempotency_conflict on Sume. How to key each script version so edits never collide.
Eleven v4 Turbo in ElevenAgents vs Sume TTS as async jobs
Eleven v4 Turbo targets live agents. Sume TTS is an async job with poll or webhook and no streaming. Which one fits a call bot and which fits produced audio.
Etsy seller listing video from one photo with Wan 3.0 on Sume
Make a short handmade-product listing video from one photo with Wan 3.0 on Sume: 2 to 30 seconds, reference inputs, an async job, and a pre-upload check.
FLUX.2 pro or flex for label text on a bottle: Sume ids
Black Forest Labs calls flex the typography and detail model and pro the affordable production model. How to compare them on Sume with one label prompt.
Food truck weekly special clip: Gemini Omni Flash 1.1, vertical, sound
Make a 3 to 10 second vertical clip of this week's special with Gemini Omni Flash 1.1 on Sume: native synced audio, 9:16, from a dish photo as a reference.
Format showcase field: judge a Format by a real output first
GET /v1/formats returns showcase, a verified sample output. Use it with description and io to pick a Sume Format before you spend credits on a run.
Gemini 3.8 Flash-Lite TTS for scratch VO vs a Sume TTS first cut
Flash-Lite TTS is Google's cheaper speech tier. If you want a scratch read before locking a script, here is how a first cut on Sume TTS 1.0 compares.
Gemini TTS 24/16/8 kHz vs Sume TTS six sample rates up to 48 kHz
Gemini 3.8 Flash TTS offers 24000, 16000 and 8000 Hz. Sume TTS 1.0 accepts six rates from 8000 to 48000 Hz. Which rate suits video, phone and web.
Gemini 3.8 Flash TTS voice library vs Sume voice selectors
Gemini 3.8 Flash TTS lists 30 curated voices plus a larger extended library. Sume picks a voice by avatar or voice id. How the two selection models differ.
Gemini API leads with Omni Flash, Veo 3.1 for specialists: and Sume?
Google's video docs recommend Gemini Omni Flash as the default and Veo 3.1 for specific needs. Sume has sume/auto or a pinned catalog id. How they line up.
Two failure channels in the Sume SDK: submit error vs failed job
A generateVideoV1 error means no job exists; a failed job means one did and billing was settled. Handle both channels in TypeScript without double-submitting.
GPT Image 2.5 background: opaque for listings, transparent
The background field takes auto, transparent or opaque on GPT Image 2.5 only. Use opaque with a white-backdrop prompt for listings, transparent for stickers.
GPT Image 2.5 Sunburst: a five-turn product photo edit loop
OpenAI positions Sunburst for tighter control over repeated edits. A five-turn loop on Sume: one source photo, one change per call, a preserve list each time.
Griffin-style follow-ups with rendered avatar clips and branching
Griffin-Lite reacts live. Until you can use it, approximate a guided conversation with a set of pre-rendered Sume avatar clips and your own branching logic.
Gym promo for Reels: three beats in 8 seconds with Gemini Omni
Write a 9:16 fitness promo for Gemini Omni with timecoded beats, a sound line and a $1.00 720p price on Sume. Includes a prompt you can paste.
House cleaning before-and-after reel with the sume-before-after Format
A cleaning company can turn a before photo and an after photo into a reveal video with one Format call. What to send, what to cap, and what it cannot prove.
Order id or payload hash? Choosing a Sume Idempotency-Key
An order-id key gives 409 idempotency_conflict when the prompt changes; adding a payload hash gives a new paid job. Choose on purpose, with Python.
Ideogram 4.5 edits: first reference is the source, 4 more allowed
On Sume, the first Ideogram 4.5 input reference is the image being edited and up to 4 more are references: 5 total. Order them or the wrong photo gets edited.
Ideogram 4.5 costs the same at 1K and 2K: pick 2K by default
Ideogram 4.5 is priced by quality tier, not size: $0.03, $0.06 or $0.22 list. On Sume you choose resolution 1K or 2K, so the size is a free choice.
Sume /v1/images returns 200 or 202: branch on the status code
A slow 4K or xhigh image request on Sume returns 202 with a job envelope, not the image body. A Python client that handles 200, 202 and 502 correctly.
Indie skincare brand launch kit: texture still, serum-drip still, reel
An indie skincare brand can make a texture still, a serum-drip still and a studio reel from one packshot with three Sume catalog Formats. What each one does.
Instagram's Reels page says 3 and 20 minutes: which one to plan for
Instagram's Reels features page says both multi-clip videos up to 3 minutes and clips adding up to 20 minutes. Plan creative for 3, tools for 20, and read both.
Interactive avatar or avatar video? A five-question test
Griffin-Lite is a research preview. Five questions tell you whether you need a live avatar or a rendered Sume Avatar 1.0 clip, and what to ship this quarter.
terminal, result_ready or status: which check ends a Sume poll loop?
Stop on terminal, fetch on result_ready, branch on status. Three fields in the Sume status payload, three different jobs, and a Python loop that uses each once.
Kling motion control: Standard 9 credits, Pro 12, Sume runs Standard
Kling lists 9 credits a second for Standard and 12 for Professional Motion Control. Sume runs the Standard queue at $0.1575 a second, in dollars.
Kling motion control still: 340 to 3,850 px edges, checked first
Kling's motion control guide sets a 340 px short edge and a 3,850 px long edge for the still. Check both in Python before a Sume job reserves money.