Updates that matter when you build

Google's Gemini 2.5 Pro text-to-speech is now live on SpicyAPI as google/gemini-2.5-pro-tts/text-to-speech.

• Input: a script of up to 5,000 characters. Cues in square brackets such as [sighs] are performed, not read aloud.
• Voice: one of 30 prebuilt voices, each with a fixed character — from bright and firm to breathy and warm.
• Delivery: an optional style (whisper, empathetic, vocalsmile, newscaster, promohype or deadpan), one of 8 English accents, and free-text instructions on pace, mood and who's listening — none of which is spoken aloud.
• Output: one WAV file, 24 kHz, 16-bit mono.
• Price: $0.00045 per second of audio returned (about $0.027 per minute), 10% below Google's list price. A hold for the longest possible clip is placed while the task runs and released once the audio is measured; failed requests are not charged.

How far it goes. We add no content filter. In our tests it read NSFW scripts in full — direct lines, narrated fiction, breathy vocal cues — without skipping words or cutting the clip short. There's no two-voice dialogue on this endpoint and no voice cloning: one voice per request, from the 30 prebuilt options.

Try Gemini 2.5 Pro TTS →

Anthropic's newest Opus, released on 22 September, is now on SpicyAPI as anthropic/claude-opus-5.5/chat, one OpenAI-compatible endpoint at /v1/chat/completions.

• Price: $3.60 per million input tokens, $0.18 cached, $4.50 cache write and $18 output, 10% below Anthropic's list price.
• Window: 1,000,000 tokens in, up to 128,000 out.
• Thinking is always on: reasoning_effort takes low, medium or high, with medium as the default.
• Tools: up to 128 definitions; tool_choice accepts auto or none, because the model rejects a forced tool call.

How far it writes. We add no content filter of our own. On launch day we sent it 24 adult-content probes across three prompt styles and it refused none. Under an adult-fiction system prompt it wrote every intimate scene in full; asked directly, it usually wrote a toned-down version instead, where Claude Opus 5 refused outright.

Try Claude Opus 5.5 →

Kling's current video lineup is now on SpicyAPI, billed per second in US dollars with no package to buy up front.

• Kling O3 (kling/o3/text-to-video, /image-to-video, /reference-to-video, /video-edit): Kling 3.0 Omni. Cast up to seven reference photos into one shot, edit a clip you already have (a 3 to 10 second source), and turn on smart multi-shot cuts and native sound, up to 4K.
• Kling 3.0 (kling/3.0/text-to-video, /image-to-video, /character-animation): the flagship, with a first and last frame, up to 4K and switchable sound. Motion Control copies the moves, expressions and gestures from a 3 to 30 second clip onto your own character.
• Kling 3.0 Turbo (kling/3.0-turbo/text-to-video, /image-to-video): every clip comes with native sound, at 720p or 1080p.
• Kling 2.6 (kling/2.6/text-to-video, /image-to-video, /character-animation): Kling's first native-audio model, 1080p in 5 or 10 seconds, with its own Motion Control.
• Kling 2.5 Turbo (kling/2.5-turbo/text-to-video, /image-to-video): budget 1080p in 5 or 10 seconds, with a closing frame, a negative prompt and guidance.
• Kling 2.1 (kling/2.1/image-to-video) and Kling 2.1 Master (kling/2.1-master/text-to-video, /image-to-video): the previous generation, from 720p image-to-video to Master's high-fidelity 1080p.
• Kling Avatar 2.0 (kling/avatar-2.0/talking-avatar): one portrait and a voice track make that person talk or sing for as long as the audio runs, up to about five minutes.

Kling reviews content itself: suggestive shots sometimes pass, anything more revealing is refused, and a refused request is never charged. For adult work, use one of our Spicy families.

Kling O3, Kling 3.0, Kling 2.5 Turbo and Kling Avatar 2.0 are priced below Kling's own per-second list price.

The review center now ranks models by how good their output looks, not by how many specs they list.

• Output quality is 85% of the Spicy Index. It combines public human-vote leaderboards (arena.ai, Artificial Analysis, Design Arena) with our own blind test: 112 prompts, every model given the same prompt, scored by three judges from different model makers who never see the model names. Specs such as resolution, duration and accepted inputs count for 15%, and resolution tops out at native 1080p.
• The blind test is checked against human votes. On a board where our judges don't agree well enough with public human-vote leaderboards, the blind test counts at a reduced weight, and a model with only blind-test evidence there is shown but not ranked. This round that applies to video, where the judges mostly see sampled frames and miss camera work. Every test report says which rule applied.
• Sound counts too. On six audio prompts we rated how clear the speech is, lip sync, emotion, and whether each sound lands on the right moment.
• 69 new reviews in 10 languages: 32 image and video models, 20 language models and 17 tools, each with our own test results and current prices.
• 106 new prompts in the arena, with 1,693 outputs to vote on.

The full rules are on the methodology page.

The Tools panel in Spicy Studio now lists every tool, grouped into From an image, From a video and People & faces, with the price on every tool that runs from the panel. Pick a tool, then choose one of your results or upload a file.

New in Studio:

• Face swap on photos comes in three levels. Standard is quick and keeps the hair and head shape of your photo, Plus matches skin tone and decides for each photo whether to move the face alone or the whole head, and Pro keeps the most facial detail.
• Head and character swap on video run in two steps. Studio first finds everyone in the clip, you pick who to replace, and only then does the swap run. If the first step finds no one, it is refunded and nothing else runs.

The first time you use a tool that works on a real person's face, talking avatar and lip sync included, Studio asks you to agree to three rules: use only your own face or the faces of people who said yes, never use a photo of anyone under 18, and never use it to pass yourself off as someone else or to mislead anyone. Your answer is remembered on that device.

Open a canvas from Canvas and it now fills the screen, with a dot grid, your shots as nodes and the controls around the edges.

• Add a node anywhere. Double-click or right-click empty space, press + in the bottom toolbar, or press N. Add an image, a video, an edit, a note or a four-shot scene, or bring in an upload, a result from My work, a template, a style or an effect.
• Build on any result. Pictures and clips have a + on the right that opens Generate from this node: make a video from a picture, edit it or give it a new look, turn a clip's last frame into a picture, or run any tool that works on it. Prompt nodes have a + on the left that adds an input: an upload, something from My work or another node on the canvas.
• Plan before you run. Connect an image node to a video node and the video starts from that image as soon as it is made. The prompt box shows the price of this step and of both together.
• Prompt in place. Select a prompt node and its prompt box opens right beside it, with the model, references, style and settings, and the price on the send button. Connected pictures are numbered, so you can type @ to point at one, for example @image2. With video models that take an end frame, choose whether two pictures are the first and last frame or references.
• One toolbar. Select, hand, library, My work, tools, undo and redo sit in the bar at the bottom; zoom, fit to screen and the mini map are in the corner. The lines between nodes and the storyboard view are switches at the top.

Nothing runs until you press generate, and every run shows its price first.

The prompt box in Spicy Studio no longer squeezes every model into four shapes and three lengths. It shows what the model you picked actually accepts.

• Aspect ratios as numbers. Every ratio the model takes is listed, from 1:1 and 4:5 through 9:16 to 21:9, each with a small frame drawn to scale. If the model can choose the shape itself or follow your photo, Auto comes first.
• Every supported length. Up to seven lengths appear as buttons; wider ranges get a slider. If the model offers Auto length, Studio shows the longest the clip can run.
• Switch models without starting over. If the new model doesn't offer your ratio or length, Studio moves to the closest one and tells you what it changed.
• Video templates you can preview. All 20 video templates now have a real sample clip that plays when you hover over the card. Each was made with MiniMax H3 from the template's own prompt, and MiniMax H3 is the model the template opens on. The template library also switches between Images and Videos.
• Style cards keep their subject when you hover, so you compare looks, not people.

Three new image families run open models with up to three of your own LoRA files per request, each with its own strength. Every image is one flat price, whatever the aspect ratio.

• Z-Image Turbo LoRA (alibaba/z-image-turbo-lora/text-to-image, /edit): Alibaba's 8-step Z-Image Turbo, a few seconds per image, nine aspect ratios at about one megapixel, plus image-to-image with a strength control.
• FLUX.1 Dev LoRA (black-forest-labs/flux-1-dev-lora/text-to-image): Black Forest Labs' FLUX.1 [dev], the base with more than a thousand community LoRAs on Hugging Face, with a guidance control.
• Qwen Image 2512 LoRA (alibaba/qwen-image-2512-lora/text-to-image, /edit): Qwen-Image-2512 at up to 1536 pixels on the long side, and Qwen-Image-Edit-2511 editing with one to three sources and edit LoRAs such as new camera angles.

A LoRA is a direct link to a .safetensors file: no upload, no allow-list, no review. All five endpoints are uncensored, and loading LoRAs adds no separate charge.

Lip sync (spicyapi/lip-sync-v1/lip-sync) now counts the seconds of the audio you send, not the seconds of the video it returns. The returned clip usually runs about half a second longer than the audio, and that extra time is no longer billed.

The rate is unchanged: $0.012 per second, rounded up to the whole second, with a 3-second minimum. A 4.5-second voice clip is now billed as 5 seconds, $0.06; counted on the returned clip, it could come to 6 seconds, $0.072. Your requests stay the same, and the same rule applies in Spicy Studio.

Spicy Studio opened on 26 September at spicyapi.ai/create. Same models and same per-run prices as the API, in a workspace made for creating rather than integrating.

• One line, then Generate. Pick Image or Video, write what you want to see, or press Tab to use the example. Each mode opens on a sensible model and you can switch at any time; your prompt comes with you. The price on the button is the most the run can cost, and nothing is charged until you press it.
• 97 looks, 4 finishes and 12 effects. A look wraps your subject in tested wording for light, colour and medium — film, editorial, anime, painting, 3D and an After Dark set — and shows the full prompt before you run. It adds nothing to the price. An effect turns one photo into a finished shot: push-ins, orbits, drone rises, rain, snow, a 16mm look.
• 150 templates in 13 groups. Each is a finished brief with the set, light and lens written; you only say who is in the frame.
• Canvas for pieces with more than one shot. Branch any frame into a video, an edit or variations, turn the last frame of a clip into the start of the next, then group takes into scenes and play them back as a storyboard. Example projects show every prompt and can be copied whole.
• Tools on every result: the eight new tools, from background removal to lip sync, sit next to Make it move and Variations.
• Enhance rewrites a one-line idea into a detailed prompt for the chosen model before it runs, billed separately at the small price shown on the button.
• Private by default. Nothing is published to a feed or gallery. Results stay in My work for 14 days, uploads are deleted after one day, and a blur switch hides results until you tap them.

You can browse, load examples and see exact prices without an account; you need one to generate, and your draft is kept while you sign up. Every run in Studio is also an API call with the same parameters, so a look you settle here moves straight into code.

The site header changed with it: Models · Studio · Rankings · Developers · Pricing. The model collections that used to sit under Explore are now under Collections in the Models menu.

Eight single-purpose endpoints went live on 26 September. Each does one job on an image or clip you already have, at a flat price. Call them like any other model, or use them in Spicy Studio, where they sit under Tools and on your results.

What decides the bill:

• Video upscaling bills on the source clip's length, rounded up to the whole second, with a 3-second minimum.
• Talking avatar and lip sync bill on the returned video, which runs as long as your audio (3-second minimum). Lip sync loops a clip that is shorter than the audio.
• Sound effects bill on the duration_seconds you ask for (1–30, default 8), not on your clip, so set it to the clip's length. It makes effects and ambience, not dialogue or music.
• Audio and video inputs must be direct links to the file, and uploading them is the most reliable route. Inputs outside the allowed length are rejected before anything is charged.

xAI released Grok 4.7 on 21 September, and xai/grok-4.7/chat has been callable here since 22 September, on the same OpenAI chat format and SSE streaming as the other Grok endpoints.

• 40% below xAI's published price on the standard tier ($2 / $0.50 / $6), and 60% below its long-context tier ($4 / $1 / $12).
• Uncensored on this route. On 22 September, nine creative-writing probes at level 4 of our five-level content ladder, asked three different ways, came back as nine complete scenes with no refusal; on 24 September every other level we ran passed too, the top one included. We add no filter of our own.
• The system role is honoured, checked on purpose, so a frame set at the top of a thread carries through it.
• Reasoning bills as output. The high default costs more than it looks; for dialogue, low or medium reads better and costs less.