Kling 4.0 video engine and Ideogram vectorization
- Kling 4.0 — new video engine
kling_4_0onPOST /videos/generations, inGET /videos/models, and on the MCPgenerate_videotool. All three modes:image_to_video(with an optional end frame),text_to_video, andtext_to_video_styled. - Durations — every whole second from
3to30. The MCP tool uses5when you omitduration; the REST API requires it, as for every engine. - Resolution and aspect ratio —
720p(default) or1080p;16:9,9:16, or1:1. Inimage_to_video, the video follows the framing of the start image. - Audio — a soundtrack is always generated, at no extra cost.
- Credits —
15per second at720p,25per second at1080p(a 5 s clip is75/125).text_to_video_styledadds the usual flat10. - Vectorization engines —
POST /images/vectorizationsand the MCPvectorize_imagetool have a new optionalengine:recraft(default) orideogram. Withideogram, the optionalmodeisdefault(detailed) orsimple. A request withoutengineworks as before. An unknown value, or amodewithoutideogram, returns422before any charge. - Vectorization credits stay at
5for every engine and mode. A cached result (0credits) now counts only when it comes from the same engine and mode.
Image uploads, Magic Edit engines, and terminal job status
- Image uploads — new. Upload a local image file and get an
image_idin three steps:POST /images/uploads/prepare(send the exactsize_bytes) →PUTthe file to the presignedupload_url(valid 1 hour) →POST /images/uploads/{image_id}/complete. JPEG, PNG, or WebP; up to 10 MB and 32 MP; at most 20 uploads awaiting completion per user. Free. Needs any one ofimages.edit,images.transform, orvideos.generate— no new scope. Theimage_idworks as the source of edits, image operations, and image-to-video. - MCP — two new free tools,
prepare_image_uploadandcomplete_image_upload, do the same. The server now has eighteen tools. - Magic Edit engines —
POST /images/editsand the MCPedit_imagetool now also acceptnano_banana_2andnano_banana_2_1, next tonano_banana_pro(default),gemini, andkontext.resolution(1K/2K/4K) works on the three Nano Banana engines, at 5 / 10 / 20 credits per output;geminiandkontextstay at a fixed resolution and 5 credits. - MCP
edit_image— new optionalresolutionandnum_imagesparameters. Before, MCP edits always ran at1K. - Edit response — the
202now echoes the effectiveengineandresolution. - Behaviour change: trained models with another engine — trained models and a
color palette in an edit work only with
nano_banana_pro. A request that names another engine together with them now returns422before any charge. Before, such a request was silently switched tonano_banana_pro. - Terminal job status for compliance blocks — an image job whose outputs all
failed the compliance pipeline (for example C2PA signing) now ends
failedwitherrordelivery_blocked_compliance: <codes>, instead of stayingprocessing. A job with some deliverable outputs iscompletedand lists the others in the newblocked_imagesfield. This applies to the REST jobGETs and the MCPget_job_statustool. - 4K outputs — C2PA signing of large (4K) outputs works again, so 4K images are delivered again.
Text-to-image engines
POST /images/generations and the MCP generate_image tool now accept seven
text-to-image engines. The default stays nano_banana_pro; a request without
engine behaves as before.- Engines —
nano_banana_pro(default),nano_banana_2_1,nano_banana_2, and new:nano_banana,imagen_4,seedream_4,flux_1_pro. Each engine has its own aspect ratios and resolutions, validated before any charge (422). GET /images/models— new. Lists the engines you can use, with aspect ratios, resolutions,num_outputslimit, credits per image, and trained-model support. Free, scopeimages.generate.- MCP —
generate_imagehas a new optionalengineparameter. This replaces the earlier note thatgenerate_imagealways uses the default engine. The new free toollist_image_enginesreturns the same content asGET /images/models. - Trained models and palettes work with
nano_banana_pro,nano_banana_2_1, andnano_banana_2. On the other engines they return422— never dropped. - Credits are the same on every engine: 5 per output at
1K, ×2 at2K, ×4 at4K. - Behaviour change for organizations restricted to trained models — they
get an empty engine list (
unavailable_reason: "organization_restricted"), anyenginereturns422, and a request must omitengineand pass at least one trained model or a palette.
Nano Banana 2.1
A new image engine,nano_banana_2_1 (Nano Banana 2.1), is available next to
nano_banana_pro and nano_banana_2.POST /images/generationsandPOST /images/img2imgacceptengine: "nano_banana_2_1". The default engine staysnano_banana_pro.- MCP — the
img2imgtool acceptsengine: "nano_banana_2_1". At this release,generate_imagehad noengineparameter and always used the default; it has one now (see Text-to-image engines above). - Same aspect ratios as
nano_banana_2(img2img also allows4:1,1:4,8:1,1:8), resolutions1K,2K, and4K, and up to 14 img2img sources. - Same credits as
nano_banana_2: 5 per output at1K, ×2 at2K, ×4 at4K. - Not available for Magic Edit (
POST /images/edits) or the MCPedit_imagetool — both return422.
Image operations
Six ways to transform an image you already have, each an asyncPOST that
submits a job plus a GET that polls it.POST /images/img2img·GET /images/img2img/{img2img_id}— transform 1–14 source images with a prompt.POST /images/variations·GET /images/variations/{variation_id}— creative variations of one image, targeted at the whole frame, a person, an object, or the scene.POST /images/resizes·GET /images/resizes/{resize_id}— change aspect ratio by outpainting rather than cropping.POST /images/upscales·GET /images/upscales/{upscale_id}— upscale to a higher resolution across four models.POST /images/background-removals·GET /images/background-removals/{background_removal_id}— cut the background to a transparent PNG.POST /images/vectorizations·GET /images/vectorizations/{vectorization_id}— trace an image to SVG.
images.transform. Img2img runs under the existing
images.edit. Existing keys created with all scopes already carry it; a
narrowed key needs the scope added in Settings → API Keys.Also in MCP. The same six operations are available as the img2img,
create_variations, resize_image, upscale_image, remove_background, and
vectorize_image tools.Pricing. Background removal costs 1 credit and vectorization 5, both 0 on a
cache hit for a source already processed. The rest bill at 5 × resolution × outputs; upscale bills by target tier times a per-model multiplier. Image Operations ·
PricingHosted image watermark decode
The content verification API now decodes both image watermark techniques itself, rather than reporting them as unchecked.- TrustMark Q — the legacy corpus, covering images marked while TrustMark was the embed vendor. This lane stays available for that corpus permanently.
- Meta PixelSeal — the current image watermark carried by newly generated Samsa images.
detected verdict still comes from
C2PA manifest verification, which is authoritative. Video watermark decode
(Meta Video Seal) is not yet available and is reported as not_checked — a
technique that was never evaluated, never a false negative. Content verificationContent verification API
A free, unauthenticated API for checking whether an image or video was made with Samsa — Samsa’s public detection mechanism under Article 50(2) of the EU AI Act. It has its own base URL, takes no API key, and spends no credits.POST /v1/public/detect— check one uploaded image or video.GET /v1/public/detect/info— machine-readable description of each marking technique and how to detect it.GET /v1/public/detect/result/{request_id}— signed PDF of a past result.GET /v1/public/detect/health— service status and version.
Trained models by name, palettes, and inline previews
Three changes that make the MCP tools easier to drive from a conversation, where a model’s name is at hand but its id is not.- Names resolve like ids.
style_id,object_ids,person_ids, andsetting_idsaccept a trained model’s name or id; a name resolves to a model visible to the connected credential. color_palette— a new parameter ongenerate_imageandedit_image, also given as a name or an id.edit_imagereached parity withgenerate_image, accepting all four trained-model reference parameters so an edit stays on-brand.get_job_statusreturns a preview. A completed raster image job now carries a downscaled inline image alongside the link to the full-resolution asset, so clients can render the result directly. Vectorization returns SVG and has no raster preview.
The Samsa public API and MCP server
The first public release. Samsa’s image and video studio — and your organization’s trained models — over a REST API and a remote MCP server.- Images —
POST /images/generationsandPOST /images/edits, composing your organization’s trained style, object, person, and setting models. - Video —
POST /videos/generationsfrom a start frame, from text, or from text styled with your models, plusGET /videos/modelsfor the engine list. - Model training —
POST /models,POST /models/prepare,POST /models/{model_id}/complete, andGET /models/{model_id}/status. - Models — list, retrieve, update, and delete at
/models. - Account —
GET /me,GET /credits, andGET /usage.
webhook_url to be notified on completion instead of
polling. Callbacks are signed, and failures retry at 5s, 30s, 2m, 15m, and 1h.MCP server. A remote Streamable-HTTP server at https://api.samsa.ai/mcp,
taking either OAuth 2.1 for interactive clients such as Claude and ChatGPT or an
API key for headless ones. Samsa API · Quickstart ·
MCP server