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
Lyria 3.5 better vocals and pronunciation: a listening checklist
Google says Lyria 3.5 improves vocals, lyrics and pronunciation. Check a Sume Music 1.0 take yourself: listen, transcribe it with STT, and compare the words.
Omni Flash or Veo 3.1 for the Gemini API? What Sume runs
Google's video docs now call Omni Flash the default over Veo 3.1. Here are Google's listed prices for both, and which one Sume's video router carries.
Should I use sume/auto for brand image work? Pin a model id instead
sume/auto picks the image family for you and never says which. For brand work where consistency matters, pin a catalog id. What the docs say and how to choose.
Gemini Omni scene extension: app, Flow, API or Sume?
Google's launch post lists four places Omni 1.1 Flash runs. Scene extension is not in all of them, and Sume's router does not expose it. Where to go instead.
AI music video with a singer close-up: still plus audio, 5 to 14.8 s
The lip-sync model in Sume's video catalog takes a still and audio of 5 to 14.8 seconds. Split a song to fit, and join the shots on a Timeline.
Which AI video models take an end frame or reference clips on Sume
Sume's catalog flags show who accepts text-to-video, an end frame, reference images, clips and audio, and who edits video. Grok is image-to-video only.
Wan 3.0 Apache-2.0: repo license line vs hosted API on Sume
The Wan 3.0 GitHub page shows an Apache-2.0 license line but does not say weights are released. What that does and does not tell you, plus the wan-3.0 id.
Lyria 3.5 tempo and duration control: Flow Music versus the Sume API
Google says Lyria 3.5 gives easier tempo and duration control in Flow Music. On Sume the same model has no duration field, so you steer length in the prompt.
Exact 16:9 sizes on GPT Image 2.5: 1280x720, then steps of 256
Only widths that are multiples of 256 give an exact 16:9 box GPT Image 2.5 accepts. 1280x720 is the smallest; 1024x576 and 1920x1080 both fail.
Tokens in a 1280x720 GPT Image 2.5 image: 106 to 3,787
A 1280x720 GPT Image 2.5 output is 106 tokens at low and 3,787 at max. Token count, dollars at $30 per 1M, and images per million tokens.
1792x1008 on GPT Image 2.5: the exact 16:9 size between 720p and 1440p
1792x1008 is a legal exact 16:9 size for GPT Image 2.5. About $0.0052 at low to $0.1850 at max on Sume, with the full quality table.
Kling 3 aspect ratios on Sume: 16:9, 9:16 and 1:1 only
kling-3 on Sume accepts 16:9, 9:16 and 1:1. For 4:3, 3:4 or 21:9 pick Seedance 2.5, Wan 3.0 or MiniMax H3 instead. A 400 means the value is not on the list.
Kling 3.0 on Sume: kling-3 is v3 Pro, motion control is separate
On Sume, kling-3 is Kling Video v3 Pro for prompt and frame video; Kling 3.0 Motion Control is a different surface billed on the driving video's seconds.
Open-weights video wave: which ids Sume lists, and which it does not
Sume's v1 video catalog lists Seedance, Wan 3.0, MiniMax H3, Kling 3 and Omni 1.1. It has no LTX id and no weights or LoRA fields. How to check before you plan.
MiniMax H3 audio references: 2 to 15 s each, 15 s combined
On Sume, MiniMax H3 audio references must be 2 to 15 seconds each, 15 seconds combined, at most 3 files, and never the only reference. Here is how to prep them.
Which Sume video ids turn audio off, and which add sound
Gemini Omni 1.1, MiniMax H3 and H3 Max always produce audio on Sume; generate_audio false is rejected. Kling 3 prices sound on and off, Wan reads a flag.
Omni Flash edit on Sume: what editing a 6 s and a 30 s clip costs
Editing a clip with Omni Flash on Sume bills per output second at the resolution you pick: $0.75 for 6 s at 720p, up to $3.75 for a 30 s hint. How it is priced.
Sonic 3.6, 3.5 and 3 cost the same on Sume TTS Router: pick by ear
All five Sume TTS Router ids share one per-character price, so model choice is a listening test, not a budget one. Here is how to run it for a few cents.
TikTok's three ad ratios and Kling 3's three: 9:16, 16:9 and 1:1
TikTok's non-Spark ad specs list 9:16, 16:9 and 1:1. Kling 3 on Sume lists exactly those three, at 720p or 1080p and 4 to 15 seconds. Here is what that fits.
Wan 3.0 reference videos: 5 clips, 15 s total, 16 fps floor
Wan 3.0 on Sume takes up to 5 reference videos, 15 seconds combined, each at 16 fps or higher. Check clip count, length and frame rate before you submit.
GPT Image 2.5: image_size wins over aspect_ratio, plus the pixel rules
Send aspect_ratio 1:1 and image_size 1280x720 to GPT Image 2.5 on Sume and the pixels win. The multiples-of-16, 3:1 and pixel-count rules, with a validator.
Which Sume image models reject input_references: five text-only rows
Five of 19 Sume image models are text-to-image only and return 400 unsupported_parameter if you send input_references. The other 14 take 5, 10 or 16 references.
Claymation look test: four Sume video models at low resolution
Test a claymation prompt on Omni Flash 1.1, Wan 3.0, H3 and Seedance 2.0 Mini for under $1.20 in drafts, then price the finals at 1080p before committing.
One narrator across a series: Sume TTS 1.0 is Sonic 3.6, pin older ids
Sume TTS 1.0 always runs on sonic-3.6. To keep a series that began on an older Sonic, use the TTS Router and pin sonic-3.5 or sonic-3 with the same voice.
Neon cyberpunk street at 1080p for 10 seconds: five Sume models
A 10-second 1080p cyberpunk street clip costs $1.40 to $2.50 on Kling 3, Omni Flash 1.1, H3 Max and Wan 3.0 on Sume; Seedance 2.5 is $14.22.
Is MiniMax H3 Max 1080p native? Sume's catalog and price
H3 Max renders natively at 480p and 768p; its 1080p is a latent refinement from 768p. On Sume 1080p is $0.20 a second, double the 768p rate.
Gemini Omni Flash 1.1 reference-to-video: IMAGE_REF tags
Use up to 10 reference images and 3 short reference videos in one Gemini Omni Flash 1.1 request on Sume, and point at them with <IMAGE_REF_0> and <VIDEO_REF_0>.
Edit an existing video with Gemini Omni Flash 1.1 over an API
Send video_url and an instruction to gemini-omni-flash-1.1 on Sume to edit a clip: no duration or aspect ratio, resolution optional, native audio on.
Gemini Omni Flash 1.1 generate_audio false: why the API refuses
gemini-omni-flash-1.1 on Sume always makes native synced audio, so generate_audio: false is rejected. Replace the sound in a Timeline instead.
Which Foundry models add C2PA and a watermark: GPT Image, FLUX.2, MAI
Microsoft's Foundry page lists the image, audio and voice models that get C2PA credentials and watermarks. What that does and does not tell you about Sume.
FLUX 3 Image combines 10 references: Sume's limits per model
BFL says FLUX 3 Image combines up to ten references. Sume lists FLUX.2, not FLUX 3: the reference ceilings per catalog model, read from the live descriptors.
GPT Image 2.5 above 2560x1440 is experimental: a safe size ladder
OpenAI marks sizes over 2560x1440 experimental on GPT Image models. A ladder of valid sizes up to 3840x2160 and a Python check for Sume's image_size.
GPT Image 2.5 quality: OpenAI defaults to auto, Sume to high
OpenAI's default quality for GPT Image 2.5 is auto; Sume's is high when omitted. Why that differs, what auto reserves on Sume, and how to pin quality.
Ideogram 4.5 edit keeps the source shape when aspect_ratio is omitted
An Ideogram 4.5 edit on Sume keeps the source shape without aspect_ratio; other edit models need auto. Which ids list auto, and the 400 if they do not.
Ideogram 4.0 has no references; 4.5 takes five. What Sume lists
Ideogram's docs say 4.0 and 3.0 accept no reference images and 4.5 accepts five. Sume lists Ideogram 4.5 and 3; how its edit and reference call works.
Nano Banana 2 takes 14 references; Sume's catalog allows 10
Google lists 14 reference images for Nano Banana 2. Sume's descriptor for it is 10. How to trim, pack or choose another model before the 400.
Sume image models that do not list the 4:5 Instagram ratio
Imagen 4, Grok Image and Higgsfield Soul do not list 4:5. Which Sume ids do, what 4:5 means in pixels, and the nearest ratio to use when a model lacks it.
Which Sume image models list a 4K resolution tier
Only Nano Banana 2 and Nano Banana Pro list a 4K tier on Sume; Imagen 4 Ultra and Ideogram 4.5 stop at 2K. What 4K costs and when it returns 202.
Seedance 2.5 token formula in Python: price any clip before you send
Seedance 2.5 bills by video tokens: width x height x seconds x 24 / 1024. A Python function that turns that into a Sume dollar quote for any size and length.
Kling 4.0's 3 to 30 seconds, split across Sume models by length
Kling 4.0 spans 3 to 30 s, and Flash 3 to 20 s. No single Sume id covers 3 to 30: Omni 3 to 10, Seedance 2.5 4 to 30, Wan 3.0 2 to 30. A lookup by seconds.
Seedance 2.0, Fast, Mini and 2.5: a 10-second 720p price ladder
Four Seedance ids are callable on Sume. A 10-second 720p clip is about $1.89 on Mini, $3.03 on Fast, $3.78 on 2.0 and $5.78 on 2.5. Limits for each.
Talking video, face swap, Recast or motion control: pick by your input
Four Sume routes make a person on video. Which one fits depends on whether you have a script, a source video, a photo of each person or a motion clip.
Omni reference limits: 10 images and 3 clips versus Veo's 3 images
On Sume, Gemini Omni takes up to 10 reference images and 3 clips of 3 seconds each; Veo 3.1 takes three images and Lite none. Build the request in Python.
Veo 3.1 1080p needs 8 seconds; Omni on Sume does 3 to 10 at 1080p
Veo 3.1 only renders 1080p and 4K at 8 seconds. Sume's Omni row takes any length from 3 to 10 seconds at 1080p, so a 5-second clip costs $0.94 billed.
Japanese and Tagalog TTS: not on MAI-Voice-2.1, tagged on Sume
MAI-Voice-2.1's 23 languages exclude Japanese and Tagalog; Sume's voice library tags ja and tl. How to set language and audition for 1 cent.
Dutch text to speech API for ads: MAI-Voice-2.1 nl-NL vs Sume nl
Dutch is on both lists: MAI-Voice-2.1 (nl-NL) and Sume's voice tags (nl). Test numbers, loanwords and the je/u register for 1 cent, then price a campaign.
Danish, Finnish, Norwegian TTS: on MAI's list, Sume tags only Swedish
MAI-Voice-2.1 lists da-DK, fi-FI, nb-NO and sv-SE. Sume's voice library tags only sv. How to test a Nordic ad read for 1 cent and keep Bokmal text consistent.
Turkish TTS API: MAI-Voice-2.1 tr-TR vs Sume tr, and the dotted I bug
Turkish is on MAI-Voice-2.1 (tr-TR) and in Sume's voice tags (tr). Upper-casing a script in code turns i into I; the failure and a one-cent test to catch it.