Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Seedream 5 Lite vs Grok Image on Sume: 1.75x price, different ratios
Seedream 5 Lite is $0.04375 per image on Sume and Grok Image is $0.025, a 1.75x gap. Their aspect ratio lists overlap on 7 of 9 and 13 values.
Song limited to 90 seconds in a 165-second Short: plan the audio
YouTube says most songs run up to 90 seconds in a 3-minute Short. Join 90 s of song to 75 s of other audio with timeline audio, then render for $0.31.
Sume SDK errors by status: which class for 401, 402, 403, 404, 409
The @sume-com/sdk error classes by HTTP status, which fields they carry, and how run helpers differ from generated operations that resolve instead of throwing.
A known-answer test for the Sume webhook signature: one body, one hex
A fixed secret, timestamp, and body give a fixed sume-v1 value. Use it to unit-test your verifier: valid, stale, empty secret, and rotation cases in Python.
Tavus Growth's 10 concurrent streams vs Sume's 4, 8 and 20 job slots
Tavus Growth allows 10 concurrent streams. Sume Pro runs 4 generation jobs at once, Startup 8, Scale 20, and queues the rest. What each limit actually counts.
Text to speech API voice id: list avatars where voice.status is ready
Sume TTS rejects voice names from other services before any credit is held. Use avatar_id or avatar_handle, or a UUID or voi_ id copied verbatim.
TikTok Spark Ads: MP4 or MOV only; non-Spark lists five containers
Spark Ads list .mp4 or .mov with no duration limit; non-Spark lists five containers and 10 minutes. Sume trim and Timeline write MP4, so both fit.
Sume TTS pace test: 1,000 characters a minute decides the cap
Sume TTS stops at 20,000 characters or 1,200 seconds. The break-even is 1,000 characters a minute; measure your voice before a long read. Max $0.95 a job.
Fabric request: image_url and avatar_handle together are not allowed
Sume's VEED Fabric 1.0 body needs exactly one visual source: image_url or avatar_id / avatar_handle, never two. Valid bodies and audio rules.
Verify a Sume video callback signature in Python, empty secret refused
A short stdlib Python check for x-sume-webhook-signature on a /v1/videos callback: HMAC SHA-256 over timestamp.raw_body, rotated sume-v1 entries accepted.
Video filter limits: 8 ops, 2,048 characters, 32 filters
Video filter takes up to 8 dim/crop ops plus a filters-only graph of 2,048 characters and 32 filters. See what is allowed, what is refused, and the free check.
Video inspect with fps 0.4 returns 24 stills from a 60-second clip
The inspect frames program caps at 24 stills and fps at 2. For a 60-second clip, fps 0.4 fills all 24 slots, 2.5 seconds apart, centered in each bin.
Gemini 3.8 TTS, v4 Turbo, Scribe v2: which can you call via Sume?
None of the three. Sume's catalog (read 2026-10-09) lists sume/tts-1.0, five Sonic router ids, sume/stt-1.0 and three music ids. What to do with their files.
YouTube stereo audio at 384 kbps in a 3-minute Short is 8.64 MB
YouTube's upload page asks for 384 kbps stereo, 128 mono, 512 for 5.1, at 48 kHz. For 180 seconds that is 8.64, 2.88 or 11.52 MB; Sume detach can write 48 kHz.
Is 1080p worth 2.5x? Seedance 2.5 12-second clip by resolution
A 12-second Seedance 2.5 clip costs $3.22 at 480p, $6.93 at 720p and $17.06 at 1080p on Sume. 1080p is 2.46x the 720p price and 5.29x 480p.
Seedance 2.5 shortest clip: what 4 seconds costs vs Seedance 2, Wan 3
The shortest Seedance 2.5 clip on Sume is 4 seconds: $1.07 at 480p, $2.31 at 720p, $5.69 at 1080p. Compared with Seedance 2 and Wan 3.0.
Timeline render: 61 seconds costs double 60 on Sume
Sume Timeline reserves $0.10 per output minute, rounded up: a 60-second render is $0.10, a 61-second render is $0.20, a 30-minute render is $3.00.
What happens at zero credits mid-batch: 402 on Sume
When a Sume balance cannot cover a generation the API returns 402 insufficient_credits. Failed jobs are refunded before capture. Cap a batch locally.
Korean captions API error: caption_hangul_text_latin_style (400)
Sume refuses Korean text on the slam, punch and tiktok-green caption styles with a 400. Use black-outline, korean-ad or another Hangul style; $0.20 a job.
Crop 1920x1080 to 9:16 with an API: width 0.3164, x 0.3418
Center-crop a 16:9 clip to vertical with Sume video filter: the crop fractions, the arithmetic, and why 1280x720 falls under TikTok's 540x960 minimum.
Poster frame at 2.5 seconds: video-frames returns source-size stills
Sume POST /v1/video-frames extracts up to 24 exact stills at source size. Always async (202); a failed instant gives a null url and the job still succeeds.
Video inspect API: 24 stills max, so 2 fps covers only 12 seconds
Sume video inspect returns at most 24 stills per call and caps fps at 2. How to size frames.fps for a 60-second clip, plus fast versus precise seek.
Timeline refuses 9 chained fades: too_many_chained_transitions
Sume Timeline 1.0 refuses more than 8 adjacent fades with too_many_chained_transitions. How transitions are validated, and the hard-cut fix.
Rendered video judders: output_fps_resamples_sources and 24 to 30 fps
Why a Sume timeline render can stutter when output.fps differs from your clips, what the output_fps_resamples_sources warning means, and the fix: omit fps.
Fast video cut API: precision keyframe and actual_start_seconds
Sume video trim has two modes. Keyframe copies the stream and can start a GOP early; exact re-encodes. Both cost $0.02. Re-base on actual_start_seconds.
1:1 square AI video on Sume: eight rows take it, Omni does not
Eight Sume video rows list 1:1 as an aspect ratio. Gemini Omni Flash 1.1 allows only 16:9 and 9:16. Cheapest square clip: 5 s at 480p is $0.31.
Video fields Sume rejects: size, seed, provider.options, audio off
Sume's /v1/videos refuses size, seed and non-empty provider.options on every model, and Omni refuses generate_audio false. What to send in their place.
Talking photo greeting: 8 seconds on Fabric 720p plus voice-over cost
An 8-second talking-photo greeting on VEED Fabric 1.0 costs $1.50 at 720p or $0.80 at 480p, plus about $0.0114 of TTS for a 240-character line.
Cheapest possible clip per Sume video model: the minimum-length floor
The cheapest legal clip on Sume ranges from $0.11 (Omni, 3 s at 360p) to $1.07 (Seedance 2.5, 4 s at 480p). Per-second rates hide the minimum-duration floor.
Claude Dynamic Workflows events vs Sume's single terminal run event
Anthropic's Oct 9 Dynamic Workflows beta streams workflow_run.* events. Sume sends one terminal webhook per run, no stream. How to fan out and track progress.
frame_images or image_url? Image-to-video fields on Sume's two APIs
Sume's /v1/videos uses frame_images and input_references; /v1/video-router/generate uses image_url, end_image_url and reference_*_urls. A field-by-field map.
Gemini 3.7 Flash now routes to 3.8: where a media client pins ids
Google auto-routes gemini-3.7-flash to gemini-3.8-flash since Oct 8. Where Sume lets you pin a model, where it routes for you, and what the job records.
Gemini 3.8 Flash TTS vs Sume TTS: break-even is 47 characters
Gemini 3.8 Flash TTS costs $0.00225 per 10 seconds of audio; Sume TTS costs $0.0475 per 1,000 characters. They tie at about 47 characters per 10 seconds.
Gemini Omni Flash 1.1 prompt length on Sume: a 20,000-character check
The Sume catalog constraint for Gemini Omni Flash 1.1 is a prompt of at most 20,000 characters, 3 to 10 seconds. A short Python preflight check before you pay.
GPT Image 2.5 at 1024: xhigh is about 3,122 tokens and max about 7,024
Sume's docs put xhigh at $0.09366 and max at $0.21072 for a 1024 image at $30 per million tokens. That implies 3,122 and 7,024 tokens, before Sume's margin.
4K image prices: FLUX.3 is 6.07x its 2K, Nano Banana 2.1 on Sume 1.33x
FLUX.3 Image costs $0.607 at 4K and $0.100 at 2K, a 6.07x jump. Nano Banana 2.1 on Sume goes from $0.15 to $0.20, a 1.33x step, and Pro is flat from 1K to 2K.
Sume hosted MCP from a CI runner: API key or OAuth?
A headless runner cannot finish the OAuth consent page. Hosted Sume MCP also takes an API key as Bearer or x-api-key. What that session can call, and the rules.
Meta's 4 GB video cap at Sume's 30-minute render limit: 17.8 Mbps
Meta ad video tops out at 4 GB. A 30-minute Sume Timeline render can average up to 17.8 Mbps and fit; a 15-minute Reels ad allows 35.6 Mbps. Table inside.
MiniMax H3 768p vs Gemini Omni Flash 720p: an 8-second clip
On Sume, an 8-second clip is $0.60 on minimax-h3 at 768p and $1.00 on gemini-omni-flash-1.1 at 720p, with $0.30 at Omni's 360p. Price and constraints.
Nano Banana 2.1 0.5K is a 512 tier; YouTube thumbnails need 640 wide
The Image API names its smallest tier 512, below YouTube's 640-pixel minimum thumbnail width. 91 test thumbnails cost $6.825 at 0.5K, $9.10 at 1K, $13.65 at 2K.
One 15-second avatar script in 9:16, 1:1 and 16:9: cost of all three
Rendering one 15 s script in three aspect ratios is three jobs: $8.28 at standard, $11.025 at plus, $24.75 at max. Aspect ratio does not change the price.
Status reads for a full Sume queue: 3,600 to 72,000
Poll every accepted job every 2 s for 20 minutes: Free makes 3,600 reads, Scale 72,000. The math, and how one field cuts it.
POST /v1/videos vs POST /v1/video-router/generate on Sume
Sume's docs recommend POST /v1/videos for new integrations; Video Router stays available and unchanged. Same catalog, same jobs, different request shape.
Python verifier for Sume webhooks: rotation header and empty secrets
A Python function that checks the sume-v1 HMAC over timestamp.raw_body, accepts either signature during rotation, and refuses an empty secret. Under 30 lines.
Seedance 2.0 at 1080p or Seedance 2.5 at 720p: 10 seconds on Sume
On Sume a 10-second Seedance 2.0 clip at 1080p costs $8.51 and a Seedance 2.5 clip at 720p costs $5.78. Seedance 2.0 at 720p is $3.78. The full grid.
Make an SRT file from Sume STT word timings in Python (7 words a cue)
Sume STT always returns words[] with word, start and end in seconds. This Python function turns the list into an SRT file, seven words a cue.
TikTok ad text is fixed white; a brand typeface must be burned in
TikTok shows ad text in white, fixed font, no links, @ or hashtags. A brand typeface must be burned into the video; what Sume's caption API can set.
VEED Fabric audio must be under 10 MB: stereo wav tops out near 56 s
Sume lists VEED Fabric 1.0 audio as Sume-hosted, under 10 MB, at most 300 seconds. A 44.1 kHz stereo wav hits 10 MB near 56 seconds; a 128 kbps mp3 never does.