Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
New-hire welcome video: 30-second avatar with auto captions
A 30-second welcome video from an Avatar 1.0 talking video plus speech captions costs $5.72 on Sume. Chain, limits and English-only caveat, read 2026-10-10.
Grok Imagine's 3 voice references vs Sume reference-audio rows
xAI's Grok Imagine 1.5 takes up to 3 voice references. Sume's Grok row takes none; Seedance, Wan 3.0 and MiniMax accept reference audio under limits.
Sume Format run image attachments: 30 images, 30 MB, 500 MB
A Sume Format run takes up to 30 images, 30 MB each and 500 MB per run, in JPEG, PNG, WebP, GIF, or AVIF. Here is what invalid_attachment means.
Runway model routers vs sume/auto: what you can steer
Runway's routers pick a model by cost, latency or quality. Sume's sume/auto takes no preference, limits clips to 3-10 s and never names the family.
SQS fair queues: MessageGroupId stops one client starving Sume jobs
Amazon SQS fair queues use MessageGroupId on standard queues as a tenant tag. How to map it to clients that share one Sume account without ordering.
TikTok Spark Ads: 10,000 per account cap, un-authorize before delete
TikTok's Spark Ads page caps an Ads Manager account at 10,000 Spark Ads and says to un-authorize a video before deleting it. A rule for batches of Sume renders.
LinkedIn video ad 5% aspect tolerance: crop a tall phone recording
A 1080x2160 phone recording is about 11% off 9:16, so LinkedIn's 5% tolerance can reject it. Crop it to 1080x1920 with Sume video filter first.
Customer text in a video prompt: input vs instruction in Sume
Put customer-supplied text in the Sume Format input object, stored as data, and keep your own directive in the instruction. Here is why that split helps.
Music teacher intro video: an avatar, cutaways and a 480p b-roll
Build a 20-second music lesson intro: one Avatar 1.0 talking video, audio detach as the spine, a Wan 3.0 cutaway and Timeline. Chain and cost, read 2026-10-10.
Grok Imagine Lite draft plus Sume upscale: what a 10 s clip totals
xAI lists Lite at $0.020 per second and 1.5 at $0.080. Add Sume video upscale at $0.009 per input second and a 10 s clip totals $0.29. Read 2026-10-10.
Notion webhooks are at-most-once: reconcile against Sume job ids
Notion webhook events are delivered at most once with up to 8 retries, and carry IDs only. A reconcile loop that keeps Sume jobs from being missed or doubled.
Arazzo 1.1 workflow for Sume: submit, poll, retry until terminal
Arazzo 1.1.0 can describe Sume's submit then poll sequence over its operationIds, with a 2 s retry action. The YAML, and what I could and could not check.
One 30-second Wan take cut into three variants with Video trim
Generate one 30-second Wan 3.0 take at 480p, then cut three variants with Video trim at $0.02 each. Chain, cost and limits, read 2026-10-10.
Gym class schedule promo: a silent Wan clip and three timed cues
A gym class schedule video on Sume: one 6-second Wan 3.0 480p clip and one caption job with three timed cues. $0.575, no speech-to-text, read 2026-10-10.
Gemini CLI 63-character tool names: Sume's longest is 45
Gemini CLI truncates MCP tool names over 63 characters. Checked against Sume's tool list, the longest becomes 45, so nothing is cut.
TikTok ad profile photo is 98x98 under 50 KB: what Sume cannot output
TikTok in-feed ads want a 98x98 px profile photo under 50 KB. No Sume route outputs that size, so generate the picture and resize it outside Sume.
Performance Max rejects MP3 and WAV: turn a music track into a video
Google Ads lists audio files as not accepted on YouTube for Performance Max. Render a Sume music track over a still with Timeline 1.0 into a 10-second-plus MP4.
Failed Format runs: a dead-letter table you can retry from
Keep a dead-letter table for failed Sume Format runs: run id, thread id, failure code, retry flag, and a new idempotency key, so retries stay safe.
Coworking day-pass promo: a price card stacked over a room video
Stack a still price card over a 6-second room video with Timeline compose: Omni Flash 720p, one Nano Banana 2.1 card, 87 cents in total (read 2026-10-10).
AsyncAPI 3.1 for your Sume webhook receiver: a 29-line spec
Describe the receiver Sume calls in AsyncAPI 3.1.0: one receive operation, two signature headers, three job events. Parsed with the AsyncAPI parser.
Higgsfield Soul on Sume: n must be 1 or 4, n=2 and n=3 return 400
Sume's Higgsfield Soul row accepts n of 1 or 4 only, rejects image_size, and offers 720p or 1080p at $0.005 or $0.0075 per image. Here is how to batch it.
VS Code mcp.json: keep the Sume API key in a password input
Use a promptString input with password true so a Sume API key never sits in mcp.json. The exact config, plus when OAuth is the better path.
Qwen Image Max vs Qwen Image on Sume: 3.75x the price, no references
On Sume, Qwen Image Max bills $0.09375 per image and is text-only, while Qwen Image bills $0.025 and takes up to 10 references. Both list 13 aspect ratios.
Veo 3.1 image_url rejected on Sume: text-only error and what to use
Sume's Veo rows refuse image, frame, reference and video inputs as 'text-to-video only here'. The refused fields, other messages, and rows that take an image.
Seedance 2.5 on Runway and on Sume: same 30 s, different pricing unit
Runway lists Seedance 2.5 up to 30 seconds from $0.20 per second. Sume bills tokens: about $2.69, $5.78 or $14.22 for 10 s at 480p, 720p, 1080p.
Hedra Character 3 vs Sume Avatar 1.0: per-second cost, 30 and 60 s
Hedra lists Character 3 at 2.5 to 6.25 cents per second; Sume Avatar 1.0 is $0.184 to $0.55 per second at 720p. Totals at 30 and 60 s (read 2026-10-10).
Grok Imagine multi-image edit: xAI says 5 sources, Sume lists 10
xAI's Imagine docs cap multi-image edits at 5 source images and 10 outputs per request. Sume's Grok Imagine row lists 10 references and one output. Port safely.
Can a Cloudflare Worker wait for a Sume video? Use a webhook
ctx.waitUntil extends a Worker for up to 30 seconds after the response. A Sume Format video typically takes 15 to 30 minutes, so use a webhook instead.
Assert the Sume usage ledger in CI: canceled job debits zero
After canceling a queued Sume job, read GET /v1/usage?job_id= and assert summary.final is true and debited_usd_micros is 0. A short Python CI check.
Cursor remote MCP has no envFile: where the Sume key goes
Cursor's envFile works for stdio servers only. For Sume's remote MCP, read the key from your shell with a headers entry, or skip keys and use OAuth.
Is there an endpoint to list all Format runs? No, build an index
Sume lists runs per Format with a cursor, but there is no GET /v1/format-runs across Formats. Here is the small run index that fills the gap, and what to store.
ElevenLabs dubbing 180-minute app limit vs 3 GB API and Sume detach
ElevenLabs dubs up to 180 minutes in the app or 3 GB via API. Sume detach takes 1,800 s of source and 900 s of output per job, so long videos need ranges.
ky retry on POST for Sume: Idempotency-Key, 40 s timeout, v2 baseUrl
ky does not retry POST by default and times out at 10 s, but Sume sync can hold 30 s. A tested ky v2 config with a stable Idempotency-Key and no 429 retries.
How long do Sume webhooks retry? About three hours vs Stripe
Sume makes up to 10 delivery attempts with exponential backoff capped at an hour, about three hours in all. Stripe retries for up to three days in live mode.
Vidu Q4 clip lengths 3 to 16 s vs every Sume video row's limits
Vidu's image-to-video page allows 3-16 seconds. Sume has no Vidu row; this table maps that range to each Sume row's minimum and maximum, and who reaches 16.
Lyria 3.5 is single-turn: what that means for Sume Music retries
Google says Lyria music is single-turn, with no iterative editing and varying results per call. On Sume that means a new take per prompt: no seed, no duration.
GitHub Actions concurrency for a Sume render job: the cancel trap
cancel-in-progress stops your workflow, not the Sume job it submitted. Group by branch, cancel via POST /v1/jobs/{id}/cancel in a final step, and cap the run.
Codex default_tools_approval_mode = writes for Sume MCP
Codex sets approval per MCP server and per tool. Which of auto, prompt, writes and approve suits Sume's read and paid tools, and what output_token_limit does.
Blur and dim a video background so captions stay readable via API
Use video-filter with a dim op and a gblur filtergraph: $0.02 per encode, a free check first, 300 s source limit. Whole-frame only. Read 2026-10-10.
A video model row missing from Sume's list: why, and how to check
Veo and Genjutsu list in the Sume catalog only where their provider route is configured. How listing works, plus a Python check that fails on a missing id.
Recover Sume jobs after a crash: match your key to GET /v1/jobs
Your worker died after submit and lost the job ids. Page GET /v1/jobs, match idempotency_key to your own keys, stop at the last page. A 29-line sample.
Wan 2.7 custom audio (2-30 s) vs Sume Wan 3.0 audio references
Alibaba's Wan 2.7 takes a custom WAV or MP3 of 2-30 seconds. Sume's Wan 3.0 accepts up to 5 reference audio clips, 15 seconds in total, with an image or video.
VS Code 1.141: where to see a Sume MCP server in Customizations
VS Code 1.141 lists MCP servers found outside mcp.json in the Customizations editor. How to confirm Sume's hosted MCP is connected, and what to check next.
Is Utopai X on Sume? It is built on MiniMax H3, which Sume lists
Utopai X runs inside Utopai's PAI platform and is post-trained on MiniMax H3. Sume does not list Utopai X; it lists minimax-h3 and minimax-h3-max.
Carrd video upload: 64 MB and H.264 only, how to prep a Sume clip
Carrd accepts uploaded video up to 64 MB on any Pro plan, and only as H.264 MP4. Make the clip with Sume and use exact trim to re-encode it as H.264.
Framer looping video: keep it under 5 MB, Framer won't compress
Framer says keep looping videos under 5 MB, uses H.264 MP4, and serves a 4K upload at full size. Render and trim a Sume clip small before you upload.
Mighty Networks native video: 4 GB for hosts, 250 MB for members
Mighty Networks lets hosts upload 4 GB videos and members 250 MB (or 25 MB if set), with English-only captions. Burn other languages into a Sume clip.
Systeme.io video hosting: 2 GB max, 100 MB recommended, fit a clip
Systeme.io lets you host video up to 2 GB but recommends 100 MB or less. Plan a Sume sales-page clip around that budget and trim what does not fit.