Models
The models behind Sume's video, image, speech, and music endpoints: ids, clip lengths, resolutions, aspect ratios, audio, references, and migration.
Start with What AI models does Sume support? Every model id, by API surface
Python clip planner for a Short: split any length into model jobs
A short Python function splits a Short's length into equal whole-second jobs within each Sume video model's min and max, then prices them with Timeline.
GPT Image 2.5 at 1024px: xhigh is 3,122 output tokens, max is 7,024
Sume's docs give GPT Image 2.5 output estimates at 1024x1024: xhigh $0.09366, max $0.21072 at $30 per million tokens, i.e. 3,122 and 7,024 tokens.
GPT Image 2.5 quality auto reserves max: set quality yourself on Sume
On Sume, GPT Image 2.5 quality auto reserves max. At 1024px the max output estimate is $0.2634 with margin versus a $0.02475 low row. Set quality explicitly.
nano-banana-2 still works but runs as 2.1: check the stored job id
Sume retired Nano Banana 2: google/nano-banana-2 runs as Nano Banana 2.1 and the job stores the 2.1 id. A Python check and what you pay.
Nano Banana Pro vs 2.1 on Sume: 1.25x at 2K, 1.875x at 1K and 4K
Nano Banana Pro costs 1.875 times Nano Banana 2.1 at 1K and 4K on Sume but only 1.25 times at 2K ($0.1875 vs $0.15). A 200-image comparison at each tier.
Haiku 5.5 cache write vs read: break-even on a Sume tool catalog
Haiku 5.5 charges $0.125 per million to write cache and $0.01 to read it. For a 12,000-token Sume tool list the cache pays back on the first reuse.
Mistral Large 4 on a 12-turn storyboard: tokens are 4% of the bill
Mistral Large 4 lists $1.36 input and $4.18 output per million tokens. A 12-turn run that makes six Wan 3.0 clips spends 4.3% of its total on tokens.
Nightly read-only MCP audit: Haiku 5.5 vs Mistral Large 4 cost
A 30-turn nightly audit over Sume's read-only MCP tools costs $0.02175 on Haiku 5.5 and $0.27615 on Mistral Large 4, about 12.7x more, on stated assumptions.
Omni Flash 1.1 is billed in tokens: 5,792 per 720p second, in dollars
Google bills Gemini Omni Flash 1.1 at 5,792 output tokens per 720p second and $17.50 per million. The math to dollars, and the Sume per-second rate.
Kling 4.0 API price: not public yet; what you can call today on Sume
No public Kling 4.0 API price was found on 2026-10-09. Sume lists a kling-3 video id and Kling 3.0 Motion Control at $0.1575 per second, 1 to 30 seconds.
Edit a video with Gemini Omni Flash on Sume: fields allowed or refused
Omni Flash 1.1 edit mode on Sume takes one video_url plus a prompt, defaults to 720p, and refuses aspect_ratio, duration and other media fields.
Gemini Omni Flash references on Sume: 10 images, 3 videos of 3 s
Omni Flash 1.1 on Sume takes up to 10 reference images and 3 reference videos of at most 3 s each, so 9 s of video. No audio reference. Tags and costs.
Nano Banana 2.1 at 8:1 and 4K: 12288x1536, 18.87 megapixels
Google's size table puts Nano Banana 2.1 8:1 at 4K at 12288x1536, about 18.87 MP, larger than a 4096x4096 square. Sume lists 8:1 and 4K at $0.20 flat.
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.
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.
Nano Banana Pro 4:5 at 1K is 928x1152, not 1080x1350: the 1.16x gap
Sume's docs say Nano Banana Pro sends 4:5 at about 928x1152 at 1K, and exact 1080x1350 comes from a post-step. That is 1.16x wide and costs $0.1875 per image.
Gemini Omni rate limit: Google lists none, Sume lists them
Google's rate-limit page has no Omni or Veo row. It lists per-project RPM, TPM and RPD plus spend caps. Sume publishes per-key limits and concurrency by plan.
Does SynthID survive captions and a 9:16 crop on Omni clips?
Google says SynthID is built to survive cropping, filters and compression. What that means for a captioned or reframed Omni clip, and what Sume claims.
GLM-5.3 reasoning cannot be turned off: cap the Sume run instead
Z.ai says GLM-5.3 always reasons, with low, high and max levels. What that means for run time and for a spend cap on a Sume agent run.
GLM 5.3 Fast or Flash: which one does Sume list?
Sume's agent model list has GLM 5.3 Flash, not GLM 5.3 Fast. What Z.ai's GLM-5.3 page says and what the API's model field accepts.
Does Google keep Omni and Veo prompts for 55 days?
Google's Gemini API page says prompts, context and outputs are kept 55 days for abuse checks. Veo files last 2 days. How that differs from a Sume request.
Hy Image 3.5 Preview: 2K or 4K? Tencent and OpenRouter differ
Tencent's launch post says up to 2K; OpenRouter's page lists resolution 1K to 4K. How to test it, and which Sume image rows list a 4K tier or pixel size.
Hy Image 3.5 Preview: n=1 and seed, against Sume's n and seed
OpenRouter lists Hy Image 3.5 Preview with n capped at 1 and a seed. Sume's Image API takes n up to 4 on most rows and returns 400 for seed. A field table.
Kling.ai headlines Kling 4.0: what an API buyer can call today
kling.ai now shows 'All-New Kling 4.0', but fal still says full launch is October. Sume lists kling-3 (4-15 s, 720p/1080p) and no 4.0 id. Prices inside.
Same Omni video twice: no seed, but a replay returns it
Omni Flash lists no temperature or seed, and Sume rejects seed on every video model. To repeat a clip, replay an idempotency key or use a video_url edit.
Veo 3.1 clip older than 2 days: it cannot be extended
Veo 3.1 extends only Veo-made clips kept 2 days. If yours expired, here is what Google's page allows and what Sume's Omni edit and new generation do.
Which video model for a 12 s 9:16 clip with end frame and sound?
Seedance 2.5, Wan 3.0, Kling 3 and MiniMax H3 all take a 12-second 9:16 clip with an end frame on Sume. Prices run from $0.90 to $6.94; here is the pick.
Hy Image 3.5 Preview is not in Sume's catalog: what to call instead
Sume's image catalog does not list Hy Image 3.5 Preview. Check with one jq call, then pick a listed model by price: Soul, gpt-image-2.5, Seedream or Flux.
Kling 4.0 API: what to generate with on Sume while you wait
Kling 4.0 was announced in September 2026. Sume's catalog lists kling-3 today: 4 to 15 seconds, $0.14 a second silent, $0.21 with sound. Request included.
Nano Banana 2.1 vs Pro from 512 to 4K: price per image on Sume
Nano Banana 2.1 runs $0.075 at 512 to $0.20 at 4K; Nano Banana Pro is $0.1875 up to 2K and $0.375 at 4K. Full tier table and a per-1,000 view.
Kling 3 stops at 15 seconds: three ways to a 30-second clip
Kling 3 on Sume takes 4 to 15 seconds. For 30 seconds, use Wan 3.0 at $3.75 (720p) or Seedance 2.5, or join two takes with Timeline. Prices side by side.
Does a vertical 9:16 Omni Flash clip cost more than 16:9?
No: on Sume a 9:16 and a 16:9 Gemini Omni Flash 1.1 clip cost the same at each resolution and length. The side-by-side numbers and how to request each ratio.
Which Sume video models stop at 15 seconds and which reach 30?
Seedance 2.5, Wan 3.0, Genjutsu and H3 Max Recast go past 15 seconds on Sume; MiniMax H3 stops at 15 and Gemini Omni Flash 1.1 at 10. The full limit table.
When Seedance 2 Mini or Fast is enough, and when to pay for 2.5
A decision guide for Seedance tiers on Sume: what Mini, Fast and 2.5 cost for the same clip, what only 2.5 can do, and which jobs belong on which tier.
Five requests Omni Flash cannot take, and what to send instead
Gemini Omni Flash 1.1 on Sume refuses clips over 10 seconds, silent clips, audio references, 1:1 and 2K or 480p output. What to send instead for each.
Swap four people in one video: H3 Max Recast bills per second
h3-max-recast takes one source video and 1 to 4 person photos, 5 to 30 seconds. Price is $0.375 or $0.5625 a second whether you swap one person or four.
Gemini Omni Flash 1.1 on Sume: spec sheet and billing in one page
Everything that ships for gemini-omni-flash-1.1 on Sume: 3 to 10 seconds, four resolutions, two ratios, always-on audio, reference limits, edit mode, billing.
sume/music-auto or pinned lyria-3.5: what changes in the job
Music Router takes sume/music-auto (Lyria 3.5 today), lyria-3.5 or lyria-3-pro. What job.model and routed_model echo back, and why the price stays the same.
Higgsfield Soul on Sume: half a cent per image, batches of 1 or 4
higgsfield-soul costs $0.005 at 720p and $0.0075 at 1080p on Sume. It takes batch sizes 1 or 4, no references. Price table and a four-image call.
Seedream 5.0 Lite search and thinking: what Sume's API passes
ByteDance describes online search and thinking for Seedream 5.0 Lite. Sume's row lists 9 ratios at about $0.04375 and empty provider.options. What that means.
Moroccan Arabic speech to text: open-source Voxtral model vs Sume STT
A Moroccan open-source stack pairs language ID with Voxtral ASR for dialect code-switching. What Sume STT offers for Arabic audio, and what it lacks.
Inworld buys Ultravox, voices move to TTS-2: pin voice and model IDs
Inworld's Ultravox deal moves built-in voices to Realtime TTS-2 at no extra cost. Why to pin voice and model ids wherever you generate speech.
FLUX 3 Image is on BFL's page: which FLUX id does Sume call?
BFL lists FLUX 3 Image, FLUX.2 Max and Klein. Sume's image router calls FLUX.2 pro and flex. Ids, prices at $0.0375 and $0.0625, ratios and edit limits.
Kling Image 3.0 lists native 4K: Sume's image resolution tiers
Kling's page lists Kling Image 3.0 with native 4K. Sume's image API takes resolution 512, 1K, 2K or 4K per model; check the catalog before asking for 4K.
Seedream 5.0 Lite 'online search': not a Sume image parameter
ByteDance says Seedream 5.0 Lite can use online search. Sume's seedream-5-lite row has no search field. What the row offers: ratios, 10 references, $0.04375.
MiniMax H3 defaults to 2K at MiniMax; on Sume the default is 768p
MiniMax's H3 page says 2K is the default. Sume's minimax-h3 runs native 480p or 768p with a 768p default; 2K and 4K are billed upscales.
Chinese and English text in images: Hy Image 3.5's claim, Sume's rows
OpenRouter says Hy Image 3.5 Preview renders Chinese and English text. Sume does not list it. How to test Chinese signage on the Sume rows that are listed.
Clef context: 65,536 on Cloudflare, 16,384 on the model card?
Cloudflare's Workers AI page lists 65,536 tokens of context for Clef; its Hugging Face card says 16,384 by default. Budget from the smaller. Cost math inside.