Models
The models behind Sume's video, image, speech, and music endpoints: ids, clip lengths, resolutions, aspect ratios, audio, references, and migration.
How long is a Lyria 3.5 song, and how do I set the length?
Lyria 3.5 makes songs of a couple of minutes, and length is set in the prompt. What Google says, what Sume rejects, and prompts that steer duration.
2K AI video generator API: MiniMax H3 at 2K on Sume
MiniMax H3 makes video up to 2K. On Sume you ask for resolution 2K on minimax-h3, an upscale of native 768p. Request, price for 10 seconds, and limits.
MiniMax H3 languages: which spoken languages does it support?
MiniMax says H3 has stable support for 11 dialogue languages and partial support for others. Sume's docs list no language set, so test yours first.
MiniMax H3 Max 1080p: a latent refinement from native 768p
MiniMax H3 Max on Sume offers 480p, 768p and 1080p. The 1080p tier is a latent refinement from native 768p. What that means for price and output checks.
What is MiniMax H3 Max? The post-trained variant, explained
MiniMax H3 Max is a variant post-trained by fal.ai on MiniMax H3 for faster generation. Its resolutions, lengths and modes, and how Sume lists it.
AI video with native stereo sound: MiniMax H3 audio through the API
MiniMax H3 generates native stereo sound in the same pass as the video. On Sume audio is always on, generate_audio false is refused, and audio steers.
MiniMax H3 open weights on Hugging Face: what is in the release
MiniMax H3 has a model card on Hugging Face: two checkpoints, a 33B-parameter Transformer, a community license and a 4-GPU serving example. What to check first.
MiniMax H3 cache and sparse attention: what the open release has
MiniMax's H3 open release ships full-attention inference only; sparse attention comes later. About 13B parameters can be cached, not loaded.
AI video editing with a text prompt: MiniMax H3 and Sume's edit path
MiniMax H3 is described as editing existing video from instructions. What the vendor says, what Sume's H3 ids accept, and the id with an edit field.
Mochi video model: what Mochi 1 is, needs and can't do
Mochi 1 preview is Genmo's Apache 2.0 open video model: 480p output, about 60 GB VRAM on one GPU, and it is not in Sume's video catalog.
Nano Banana 2 Lite API: what Google shipped and what Sume lists
Nano Banana 2 Lite is Google's gemini-3.1-flash-lite-image model. What Google says it does, and which Nano Banana models Sume's image API lists today.
Nano Banana Pro custom aspect ratio: what Sume does
Nano Banana Pro on Sume takes 11 native aspect ratios. A custom ratio or WxH size is snapped to the nearest one. Which, and how to check.
Nano Banana Pro vs Nano Banana 2: which id to send on Sume
Nano Banana Pro and Nano Banana 2 share tiers and 10 references on Sume; they differ in extra aspect ratios and in which tiers change the price.
Nano Banana quality parameter: why quality returns 400
Sending quality to Nano Banana 2 or Pro on Sume returns 400 unsupported_parameter. What quality does elsewhere, and what to change instead.
Open source AI video generator with audio: LTX-2 and MiniMax H3
Two open-weights video models make sound with the picture: LTX-2 and MiniMax H3. What each page says, and how a hosted API reports audio.
Open-source video model vs API: which to use for Wan-class video
Weights you run yourself, or a hosted video API? What the choice changes for cost, setup and limits, with Wan 3.0's per-second API price as the worked example.
Qwen Image 2.1 API: open weights, and what Sume lists
Qwen-Image-2.1 is an open-source 7B image model on Hugging Face with no inference provider. Sume lists qwen-image and qwen-image-max instead.
Qwen Image 3.0: does Sume list it? A catalog check
Sume lists qwen/qwen-image and qwen/qwen-image-max, and nothing named 3.0. How to check the live catalog for a new Qwen id and what an unknown id returns.
Qwen Image license: is Qwen-Image open source under Apache 2.0?
The Qwen-Image model card says the weights are Apache 2.0 and were released on 2025-08-04. What that covers, and how the hosted qwen ids on Sume differ.
Qwen Image Max API on Sume: text-only, and how it differs
qwen/qwen-image-max on Sume is text-to-image only and rejects reference images with a 400. Price, ratios, a working call, and how it differs from qwen-image.
Recraft V4.1 API on Sume: check the catalog for the id
Sume lists recraft/recraft-v4, not a V4.1 id. How to check the live catalog for a newer Recraft, and what the listed model takes: text only, webp, $0.05.
Recraft V4 SVG: can you get vector output through Sume?
Recraft says V4 can make editable SVG. Sume's recraft/recraft-v4 advertises webp output only. What that means, plus a working call.
Does generate_audio change Seedance 2.0's price?
No. fal says Seedance 2.0 audio costs the same on or off, and Sume's Seedance price has no audio input: seedance-2 reserves about $0.38 per 720p 16:9 second.
Seedance 2.0 max length: 15 seconds, and how to go longer
Seedance 2.0 clips run 4 to 15 seconds on Sume and on fal. For 16 to 30 seconds use seedance-2.5 ($17.34 at 720p) or chain clips with first_frame.
What is Seedance 2.0 Mini, and where can you run it?
Seedance 2.0 Mini is a Seedance 2.0 model that CapCut and fal list. On Sume it is the id seedance-2-mini, 4–15 s clips, about $0.19 per 720p second.
Does a reference video change Seedance 2.0's price?
Yes. A reference video adds 15 assumed input seconds at 0.6× on Sume, so a 10 s 720p clip on seedance-2 reserves $5.67, not $3.78.
Seedance 2.0 resolution in pixels: 480p, 720p and 1080p by ratio
Seedance 2.0 720p at 16:9 is 1280×720 and 1080p is 1920×1080. The full pixel table for six aspect ratios, as fal lists it and Sume prices it, with no 4K.
Seedance 2.5 3D blockout camera control: clay renders as a reference
Seed describes clay-render referencing, a textureless 3D model that blocks camera and pose. Sume has no clay mode; you send the render as a video reference.
Seedance 2.5 aspect ratios: 9:16 vertical, 21:9 and the rest
Seedance 2.5 on Sume accepts 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. Set aspect_ratio explicitly; 3:2 is rejected. Ratios, errors and the price effect.
Seedance 2.5 audio reference: how to send an audio clip
Seedance 2.5 accepts an audio clip as a reference, even without an image or video. The audio_url request on Sume, the vendor's file limits and what to expect.
Seedance 2.5 image to video: first frame and last frame
Send a first_frame, or a first_frame and last_frame, with seedance-2.5 on Sume. Rules for frame_images, how it differs from input_references and what to set.
Seedance 2.5 omni reference: how @image, @video and @audio tags work
Seedance 2.5 prompts point at references with tags such as @Image 1. Sume rewrites lowercase tags like @image1 to the casing the model expects. Examples.
Seedance 2.5 50 references: how many can you use on Sume?
BytePlus lists 50 Seedance 2.5 references (30 images, 10 videos, 10 audio clips). Sume's API accepts 12 input_references per request for seedance-2.5. Compared.
When was Seedance 2.5 released, and where can you use it?
ByteDance Seed launched Seedance 2.5 on 2026-07-31 in its own apps. The model id seedance-2.5 is in Sume's video catalog. Dates and access, read 2026-09-29.
Seedance 2.5 video editing: green screen, camera angle, timestamps
ByteDance says Seedance 2.5 edits video by timestamp, incl. green screen and camera angle. What BytePlus documents for edits and what Sume's API lists today.
Seedance 2.5 vs 2.0: what changed in duration and references
Seedance 2.5 raises the clip to 30 seconds and references to 50 in BytePlus's docs. The 2.0 vs 2.5 table and how Sume's own limits differ, read 2026-09-29.
Seedance 2.5 vs Gemini Omni: length, references and editing
Seedance 2.5 and Gemini Omni Flash compared on published specs: 30 vs 10 seconds, reference rules, extension and edit mode. Ids on Sume and on Google's page.
Seedance 2.5 vs Kling 3.0: durations, references and audio
Seedance 2.5 and Kling 3.0 compared on the specs their own pages and Sume's catalog publish: length, resolution, references, audio. No winner, just the table.
Seedream 5.0 Pro API: Sume lists 5.0 Lite, not Pro
Sume's catalog has bytedance-seed/seedream-5-lite, 4.5 and 4.0, and no Seedream 5.0 Pro id. What a Pro id returns and which listed model to call instead.
Cartesia Sonic 3 languages: 44 in Sonic 3.6, how to set one
Cartesia says Sonic 3.6 speaks 44 languages, including new Odia and Urdu. How its language and locale fields work, and Sume's language field.
Stable Audio 3 open weights: run them yourself or call a hosted API
Stable Audio 3 Small SFX, Small and Medium are open-weight; Large is on Stability's API. Sume's music route lists Lyria ids only, so weights mean self-hosting.
Stable Audio 3 max length: six minutes, and how to reach it on Sume
Stability says Stable Audio 3 Medium and Large pass six minutes. Sume's Lyria music runs up to a few minutes; join separate takes for more with Timeline audio.
Text to speech API with emotion, speed and volume controls
Sume's text to speech API takes generation_config: speed 0.6 to 1.5, volume 0.5 to 2.0 and an optional emotion guide of up to 64 characters.
Text to speech sample rate: which output_format to pick
Sume's TTS output_format takes mp3, wav or raw, six sample rates from 8000 to 48000 Hz and four PCM encodings. The default, and when to send wav.
Veo 3 shutdown: which Gemini API models retired and what replaces them
Google retired veo-3.0-generate-001, veo-3.0-fast-generate-001 and veo-2.0-generate-001 on 2026-06-30. Its listed replacements are the Veo 3.1 preview models.
Veo 3.1 4K API: 8 seconds only, and 4K options on Sume
Veo 3.1 makes 4K video only at 8 seconds, and not on the Lite model. On Sume, gemini-omni-flash-1.1 lists 4K for 3 to 10 seconds.
Veo 3.1 API: how to call it, and what Sume lists instead
The Veo 3.1 API is Google's, called through the Gemini API with preview model codes. Sume does not list Veo; here are the catalog models with the same inputs.
Veo 3.1 extend video, and how to chain clips on Sume
Veo 3.1 extends only videos Veo made, 7 seconds at a time. Sume has no extend call, so you chain clips from a last frame or pick a model with a longer limit.