Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Keyframe or exact trim for a TikTok ad: which keeps 516 kbps?
Sume video-trim exact re-encodes with libx264; keyframe copies the stream. You cannot set bitrate, so measure size / duration against TikTok's 516 kbps floor.
MAI-Voice-2.1 and MAI-Transcribe-2: what Sume lists for voice and STT
Microsoft AI's news page names MAI-Voice-2.1 and MAI-Transcribe-2. Sume lists neither; it lists Sonic TTS (38 micros a character) and STT at $0.01 a minute.
MiniMax H3 duration errors: at most 15 s, at least 5 s, Recast 30 s
Video Router refuses a MiniMax H3 job over 15 s or under 5 s with a named message, and Recast has its own 5-30 s rule. Messages, limits and a pre-check.
nano-banana-2 still works on Sume: it runs as Nano Banana 2.1 at 10c
Nano Banana 2 is retired on Sume. The ids nano-banana-2 and google/nano-banana-2 still work and run as Nano Banana 2.1, and the job stores the 2.1 id.
Nano Banana 2 is retired on Sume: the old id still works, as 2.1
google/nano-banana-2 and nano-banana-2 still work on Sume and run as Nano Banana 2.1. The job stores the 2.1 id, and the catalog lists only 2.1.
A 20-second single shot: only Seedance 2.5 and Wan 3.0 take it
Only two Sume video models accept a 20 s request: Seedance 2.5 ($11.56 at 720p) and Wan 3.0 ($2.50). Prices at all three resolutions.
At 768p, H3 Max undercuts H3 once references pass the free five
At 768p H3 Max lists $0.02 more per second than H3 but has no per-image fee. See at which image counts and lengths H3 Max becomes the cheaper row on Sume.
Omni 4K costs 3x the 720p price and 10x 360p: the ladder on Sume
On Gemini Omni Flash 1.1 the four resolutions price at 1x, 3.33x, 5x and 10x of 360p. A table of ratios for 8-second clips on Sume.
Omni 9:16 and 16:9 cost the same: aspect ratio is not a price lever
Gemini Omni Flash 1.1 prices by seconds and resolution only. A vertical and a landscape clip of the same length cost the same on Sume. Table inside.
Omni Flash on /v1/videos vs /v1/video-router: the field map
The same Gemini Omni Flash 1.1 runs on /v1/videos and /v1/video-router/generate with different fields. Field map for frame_images, input_references, video_url.
Omni resolution values: 4K, lowercase 4k, and what is rejected
Gemini Omni Flash 1.1 on Sume takes 360p, 720p, 1080p and 4K. The catalog also accepts lowercase 4k as an alias. Which strings fail, and the price of each.
Omni video references: three clips of 3 s, so 9 s of footage at most
Gemini Omni Flash 1.1 takes up to 3 reference videos of 3 s each and 10 reference images. The tags are VIDEO_REF_0 to 2; the price is unchanged.
Poll or callback for 1,000 video jobs: request counts at 30 seconds
At a 30-second poll, 1,000 video jobs make 4,000 to 20,000 status requests by job time. A callback_url cuts that to 1,000 deliveries plus a sweep.
Polling a 5-minute video job: 150 GETs at 2 s, or 10 at 30 s
A five-minute Sume video job costs 150 status reads at a 2 s interval and 10 at 30 s. Poll /v1/videos with a loop that handles every status.
Python asyncio price check: one Omni Flash clip at four resolutions
A runnable Python script submits a 3-second Omni Flash clip at 360p, 720p, 1080p and 4K together and prints usage.cost. Expect $2.20 in total.
Recast prompt limit is 2,000 characters: what to write, what it costs
The H3 Max Recast prompt is optional and capped at 2,000 characters. See what to put in it for a person swap and why the length never changes the price on Sume.
Recast the 9:16 and 16:9 cuts of one ad: same price, no ratio field
H3 Max Recast takes no aspect_ratio; output follows the source. A 15-second 9:16 cut and 16:9 cut are $5.63 each at 768p, $11.26 for the pair on Sume.
Same ad Format, new model: log three receipt fields on every run
To keep an ad format steady when the model changes, pin the Format, treat model as the orchestrator only, and log version, model and schema.
Seedance 2 Mini vs Fast: a steady 1.6x price over 500 clips
Seedance-2-fast costs about 1.6x seedance-2-mini on Sume at every resolution: $210 more over 500 clips at 480p, $450 at 720p, $1,020 at 1080p.
Sonic 3.6 vs 3.5 vs 3 on Sume TTS: same price, so pin the id
All Sonic ids in Sume's TTS router list 38 micro-dollars per character and a 20,000-character cap. sonic-latest is an alias for 3.6. What to pin and why.
Sume's 27 catalog Formats: slugs, 404 rule and the io profile
Sume lists 27 catalog Formats at the sume handle. An unlisted slug gives 404. Read description and io before you run one, then fork it to change it.
Sume Format list empty or 404? Check which key you are using first
A missing Format is usually the wrong key: a personal key cannot see team Formats, and 403 workspace_key_required means the right team with the wrong key.
Sume webhook retries: 10 attempts, 30 s apart, a 4.5 minute window
Sume retries a job webhook up to 10 times, 30 s apart by default. That is 270 s between first and last attempt, 370 s at worst. What to do after.
Talking clip by length: 4, 15, 60 and 300 seconds, Avatar or Fabric
Under 4 seconds Avatar Video rejects; 4 to 60 s it fits; 60 to 300 s needs VEED Fabric with your own audio. Prices at each length for both routes on Sume.
TikTok 9:16 ad from Omni Flash: 5 to 10 second clips priced
TikTok recommends 9:16 for non-Spark video ads. Gemini Omni Flash 1.1 makes 9:16 clips of 3 to 10 s on Sume: $0.63 at 5 s and 720p, $1.25 at 10 s.
TikTok Spark Ads have no duration limit; Timeline stops at 30 minutes
TikTok Spark Ads accept mp4 or mov with no duration restriction. Sume Timeline 1.0 renders up to 1,800 s at $0.10 a ceil minute: 15 min is $1.50, 30 min $3.00.
Timeline bills 61 s as two minutes: the 60 s cut that halves the fee
Sume Timeline charges $0.10 per output minute rounded up, so 60 s costs $0.10 and 61 s costs $0.20. How to plan 30 s shots around it.
Timeline plan needs no Idempotency-Key; render does: a wrapper
POST /v1/timeline-1.0/plan is free and takes no Idempotency-Key. Render is paid and requires one. A wrapper plans first, then renders with a derived key.
Transcript from a video on Sume: inspect transcribe or detach first?
Sume can transcribe a video through video inspect (1,800 s limit, compute plus $0.01 per minute) or from a detached 16 kHz mono wav. How to choose.
Sume TTS price steps: 210, 421 and 631 characters per job
TTS on Sume bills 47.5 micro-dollars per character, rounded up to a cent per job, so the price jumps at 211, 422 and 632 characters. Full step table inside.
verifyWebhook toleranceSeconds: 300 by default, and 0 turns replay off
In the Sume SDK, verifyWebhook rejects deliveries older than 300 seconds by default. toleranceSeconds 0 skips that check. When each setting is right.
Wan 3.0 reference limits: 10 images, 5 videos and 5 audio tracks
Sume lists Wan 3.0 references as up to 10 images, 5 videos (15 s total) and 5 audio tracks (15 s total). Sume docs do not claim 50 references.
Is a 1:1 AI clip cheaper than 9:16? Only at 480p, by about 4%
On Seedance rows an 8 s 1:1 clip is a few cents cheaper at 480p ($2.06 vs $2.15 on 2.5) and equal at 720p and 1080p. Wan, Kling and MiniMax have one price.
Crop a 21:9 video to 9:16, 4:5, 1:1 and 16:9: fraction table
From a 21:9 source, a full-height crop is 0.2411 wide for 9:16, 0.3429 for 4:5, 0.4286 for 1:1, 0.7619 for 16:9. $0.02 per crop; the check is free.
3:2 photo ratio image API: GPT Image 2.5 does not list it on Sume
3:2 is the classic camera ratio, yet GPT Image 2.5 and GPT Image 2 do not list it on Sume. Fourteen other rows do. Table of rows and the crop workaround.
Vidu Q4 takes 3 audio references; which Sume video models take one?
Sume does not list Vidu Q4 Preview. Seedance 2.x, Wan 3.0 and MiniMax H3 accept reference audio; Gemini Omni Flash 1.1 does not. Prices for 15 seconds.
A 52-minute podcast on Sume: the 1800-second limit and the split
Sume trim and inspect refuse sources over 1800 seconds. A 52-minute episode must be split into two 26-minute parts first; six clips then cost about $2.50.
Agent turns a product page into a clip: crawl_scrape, generate_video
On Sume MCP, crawl_scrape is a read tool that works with Write off. generate_video is paid and needs Write or a key. Split the job in two sessions to cap spend.
AI avatar video aspect ratios: 5 options, 720p only on Sume
Sume Avatar 1.0 talking videos support 1:1, 3:4, 9:16, 4:3 and 16:9 at 720p. How to pick a ratio for Reels, ads, feed posts and web pages.
Backend job that needs an agent: Agent Completions or hosted MCP?
Agent Completions runs Sume's agent and returns a 202 receipt with a required spend cap. Hosted MCP is for a client whose own model calls Sume tools.
Haiku 5.5 in Cursor with Sume MCP: a 150k request costs 8x a 90k one
Cursor adds Claude Haiku 5.5 from Settings > Models. Its price steps up above 100k input tokens: 90k costs $0.009 and 150k costs $0.075 before output.
Will an AI video model lip-sync a voice-over added afterwards?
No: Sume's docs say video models do not lip-sync to generated speech or a later voice-over. A talking face needs a still plus audio, or an avatar video.
Does a Music 1.0 bed avoid Content ID on a Short? What is promised
YouTube says Audio Library songs get no Content ID claim; a Short over a minute with a claim is blocked. Sume's Music 1.0 is $0.125 and makes no such promise.
Edited one ad hook and resent the bulk run? Same key gives 409
Reusing an Idempotency-Key with a changed Sume bulk-run payload returns 409 idempotency_conflict. Send only the edited hook under a new key.
Five caption looks on one clip: one transcription, four restyles
Test five caption styles on one video for $1.00: the first job transcribes, four source_caption_id restyles reuse the word timings. Request bodies included.
How many minutes of speech fit in one 20,000-character TTS request?
Sume TTS caps a request at 20,000 characters. At 800 to 1,200 characters per minute that is roughly 17 to 25 minutes of audio and costs 95 cents.
Hy Image 3.5 lists 13 ratios; GPT Image 2.5 on Sume lacks six
OpenRouter's Hy Image 3.5 Preview ratio list has six ratios GPT Image 2.5 does not offer on Sume: 1:2, 1:4, 2:1, 2:3, 3:2 and 4:1. Which Sume rows list them.
Image API returned 202 after 30 seconds: a Python poll that finishes
POST /v1/images waits 30 seconds, then returns 202 with a job envelope for slow high-quality runs. A 25-line Python script that handles both 200 and 202.