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
GPT Image 2.5 cached input price: does Sume pass it on?
OpenAI lists cached input at $1.25 text and $2 image per million tokens for Sunburst. Sume's docs give estimates and reserved cost, not a cache discount.
Image quality default on Sume: low on Image 1.0, high on GPT 2.5
Image 1.0 defaults quality to low; ChatGPT Image 2.5 on POST /v1/images defaults to high when omitted. Values, and when to raise quality.
How many reference images? 10 on Image 1.0, 16 on GPT Image 2.5
Image 1.0 image_urls takes 1 to 10 public HTTPS URLs; ChatGPT Image 2.5 takes up to 16 references. Text-only models reject references.
gpt-image-2 or openai/gpt-image-2? Model ids on Sume's Image API
Sume accepts bare Image Router ids such as gpt-image-2 and nano-banana-2 as aliases for openai/gpt-image-2 style ids. The catalog lists org/slug ids.
FLUX video edit API: 720p downscale, and video_url edits on Sume
BFL's edit-a-video endpoint downscales sources above 720p. Sume lists no FLUX video; its edit path is video_url with gemini-omni-flash-1.1 only.
Gemini 3.5 Transcribe API limits: 1 hour vs Sume STT's 10 minutes
Gemini 3.5 Transcribe takes up to 1 hour per request, 30 minutes with diarization or word timestamps. Sume STT takes 600 seconds; split longer audio.
Gemini API speaker diarization: 8 speakers; Sume STT has no labels
Gemini 3.5 Transcribe diarizes up to 8 speakers, 3+ experimental, within 30 minutes per request. Sume STT returns word timings but no speaker labels.
Gemini Omni video editing: the EEA, UK and under-10-second rules
Google says editing or extending uploaded videos with Omni is unavailable in the EEA, Switzerland and the UK, and uploads must be under 10 seconds.
Gemini Omni extend video: Google's 40 s cap vs Sume's modes
Google's Gemini Omni can extend a video by 3-10 s up to 40 s. Sume's Omni row has no extend mode; here is what it offers instead.
Gemini TTS API: which engine Sume's TTS routes to
Gemini 3.8 TTS is not in Sume's TTS Router, which lists Cartesia Sonic ids only. What Sume's TTS takes for voice, engine and length, with a checklist.
Genjutsu Object Swap API: Sume serves Motion Transfer only
Higgsfield pairs Motion Transfer with Object Swap in Genjutsu. Sume's Genjutsu row serves Motion Transfer only; Object Swap is rejected by name.
Grok Imagine video editing: 8.7 s on xAI, Omni edit on Sume
xAI caps Grok Imagine video edits at 8.7 seconds. Sume's Grok row has no video-to-video; edits go through gemini-omni-flash-1.1 with a video_url.
Grok Imagine video sound: xAI audio vs Sume's silent row
xAI's Grok Imagine video makes audio unless generate_audio is False. Sume's grok-imagine-video-1.5 row lists audio: false and rejects generate_audio.
Hy-Image-3.5-Preview API limits vs Sume's image limits
Tencent's Hy-Image-3.5-Preview takes 256 to 8192 px edges, up to 4096x4096 and 20 references. Sume lists no Hy Image model; its own limits are below.
Ideogram 4.5 Precise Edit API: mask and reference limits
Ideogram 4.5 Precise Edit takes 4 reference images, 3 when a mask is sent. Sume lists Ideogram V3, and mask_url there is only for ChatGPT Image 2.5.
Kling 2.1 API sunset: 410 Gone, and the kling-3 id on Sume
Krea retired Kling 1.0 through 2.1: endpoints return 410 Gone after Sep 15, 2026. Sume lists kling-3 in its video catalog; here is what to change.
MiniMax H3 Max Turbo API: what Sume lists today
Krea added MiniMax H3 Max Turbo on Sep 9. Sume's video catalog lists minimax-h3 and minimax-h3-max, not a Turbo id. What each listed id accepts.
MiniMax H3 Max Turbo price: fal's rates, and the id Sume lists
fal prices H3 Max Turbo separately from H3 Max. Sume's docs list only minimax-h3-max; how to check the catalog, and what that id bills for 5 seconds.
MiniMax music API closed to new users: the Sume Music Router route
MiniMax stopped offering its paid Music and Lyrics APIs to new users on Aug 20, 2026. Sume's Music Router, sume/music-auto, is open: one POST, flat price.
Nano Banana Pro aspect ratio: auto keeps the reference shape
Freepik's changelog says Nano Banana Pro accepts aspect_ratio auto to follow reference images; omitting it stays 1:1. On Sume, omitting is also not auto.
Recraft V4.1 Flash API: Sume lists V4, webp, text-only
Recraft V4.1 Flash is live in Recraft Studio and its API. Sume lists recraft/recraft-v4, not a Flash id: webp output, text-to-image only, no references.
Medical transcription API: what Sume STT does and doesn't
Sume has one speech-to-text route, sume/stt-1.0, with no medical model option. What it takes and returns, so you can decide for clinical audio.
Seedream 5.0 pro layer decomposition vs a Sume Seedream result
Seedream 5.0 pro can split one image into a base plus up to 16 PNG layers. Sume lists Seedream 5.0 Lite, and results are one Sume-hosted URL per image.
Suno v6 audio and video input vs Sume: text and one image
Suno v6 says you can create with text, audio, images and video. Sume's music request takes a text prompt and one optional image; audio and video are not inputs.
Wan 3.0 on Runway, or the wan-3.0 id as an API job on Sume
Runway's changelog lists Wan 3.0 in tool mode and workflows on paid plans. On Sume, wan-3.0 is a model id you call from code and poll as a job.
Vidu API: S2-Avatar live voice vs Sume avatar video jobs
Vidu S2-Avatar is a real-time voice model. Sume has no live session: you submit a script and a scene photo to an avatar job and fetch the finished video.
Vidu S2-Editing: live stream edits vs a Sume Omni clip edit
Vidu S2-Editing edits an incoming video stream in real time. Sume edits one finished clip with a prompt through Gemini Omni video_to_video, as an async job.
PixVerse API and R2 world model: Sume has no live session
PixVerse R2 is a persistent real-time world. Sume has no live session or action controls: video jobs return a clip you poll for or get by webhook.
Kling 4.0 30 seconds: Sume lists kling-3 at 4 to 15
Kling 4.0 announces 3 to 30 second generation. Sume lists kling-3 at 4 to 15 seconds; seedance-2.5 and wan-3.0 are the ids that reach 30 seconds.
Kling 4.0 Omni Reference: 15 inputs; kling-3 on Sume takes none
Kling 4.0 Omni Reference allows up to 15 combined references. On Sume, kling-3 takes none; wan-3.0 and minimax-h3 list image, video and audio caps.
Kling 4.0 multiple keyframes: ten, versus Sume's two frames
Kling 4.0 announces up to 10 keyframes. Sume's video routes take a first and last frame; longer chains mean separate clips cut and joined.
Kling 4.0 video editing: five video inputs, Sume edits one
Kling 4.0 editing takes up to five video inputs. Sume's Omni edit route takes one video_url and cannot combine it with references.
Kling 4.0 prompt length: 8,000 tokens vs Omni's 20,000 characters
Kling 4.0 allows prompts up to 8,000 tokens. Sume documents a 20,000-character prompt cap for gemini-omni-flash-1.1; tokens and characters differ.
Gemini Omni reference video length: 3 seconds per clip on Sume
On Sume, each Gemini Omni Flash reference video can be at most 3 seconds, with up to 3 clips. Trim longer footage first, then address clips as VIDEO_REF tags.
LTX 2.5 API: check the Sume catalog before you send a model id
LTX-2.5 has its own API and open weights. On Sume, send only ids that GET /v1/videos/models lists; an unknown id returns model_not_found. How to check and pick.
Seedance 2.5 reference limit: BytePlus says 50, Sume varies by model
BytePlus allows up to 50 multimodal references for one Seedance 2.5 clip. Sume documents limits per model, so read the catalog for the id you send. Examples.
Grok Imagine image edit limit: xAI 5 sources, Sume catalog
xAI's Aug 28 release notes raise image editing to 5 source images. On Sume, Grok is edit-capable with n of 1; read input_references per model.
Nano Banana Google Search grounding: not a Sume parameter
Google documents Search grounding for its Gemini image models. Sume lists Nano Banana 2 but no grounding field, and unlisted fields return 400.
Nano Banana thinking mode and thought images on Sume
Google's Gemini 3 image models use thinking mode and uncharged thought images. Sume lists no thinking field, and you pay only for completed images.
Nano Banana 14 reference images vs Sume's 10 input_references
Google says Gemini 3 image models mix up to 14 reference images. Sume's Nano Banana models take up to 10 input_references; ChatGPT Image 2.5 takes 16.
Odia and Urdu text to speech API: the language codes
Cartesia's Sonic 3.6 adds Odia (`or`) and Urdu (`ur`). On Sume, send the code in `language`, pin `sonic-3.6`, and check the voice before you pay.
Lyria RealTime API: streaming music vs a Sume file job
Lyria RealTime streams steerable instrumental music over a WebSocket. Sume's Music Router returns a finished audio file per job. Here is how they differ.
Lyria 3 Clip 30-second music API: ids Sume lists
Lyria 3 Clip always makes 30 seconds. Sume does not list it; ask for a 30-second track in the prompt with sume/music-auto, lyria-3.5 or lyria-3-pro.
Kazakh speech to text API: Nova-3 and Sume language_code
Deepgram added Kazakh to Nova-3 on Sep 3, 2026. Sume STT 1.0 takes an optional language_code hint, or auto-detect. How to send audio and check the output.
TikTok Symphony with Seedance 2.5: 30-second AI video ads
TikTok Symphony now generates up to 30 seconds with Seedance 2.5, in select markets. The same 30-second length is available by API on Sume as seedance-2.5.
Muse Image API: how to check what Sume's image catalog lists
Runway lists a muse_image model at 1 credit per image. For Sume, read the live image catalog instead of assuming; endpoint pricing lines are what you pay.
HeyGen reference to video: 9 images, 3 videos, 3 audio
HeyGen's heygen-video-1 reference mode takes 9 images, 3 videos and 3 audio files, 12 in total. Here is how Sume's reference limits compare.
HeyGen text to video API: 5-15 s, 768p vs Sume durations
heygen-video-1 makes 5-15 second clips at 480p or 768p from a 5,000-character prompt. Sume lists durations and resolutions per model in its catalog.