News & Changelog
Product updates, new models, and pricing changes. Pricing changes are tagged Pricing. The live source of truth for models and prices is GET /v1/chat/models.
Something not working right now? Check live status and incidents. This page is the record of planned changes — announcements, new models and pricing.
X (Twitter) gateway: from Nov 18, post reads are 450 AP and search is 4,500 AP (X now bills expanded objects)
From 2026-11-22, X bills every object returned through
X API pricing (docs.x.com) ↗expansions(authors, media, quoted posts) at the same pay-per-use rate as a direct lookup: Post and Media $0.005, User $0.010 each. Our post read returns the author and up to four attachments in one call, and search returns up to ten authors, so both now cost us several times more upstream. From 2026-11-18 05:00 UTC:GET /v1/x/status/:idgoes from 100 AP to 450 AP per upstream fetch (one flat price, whether or not the post has images), andGET /v1/x/searchgoes from 2,250 AP to 4,500 AP per successful call.GET /v1/x/user/:username(100 AP) andGET /v1/x/timeline/:username(2,250 AP) stay the same. Cache hits and failed fetches stay free, and account-ownership verification stays free. Sources: X API pricing https://docs.x.com/x-api/getting-started/pricing and expansions https://docs.x.com/x-api/fundamentals/expansions. Published 12 days ahead, per our price-revision rule (increases get 7 days’ notice; only HTTP 200 is billed). Agent-readable spec: https://api.aicu.ai/skills/xA French national university uses AICU API for its hackathon: Arts et Métiers IAckaton
Students of Arts et Métiers (Angers, France) build startups in five days at the IAckaton, from 5 to 9 November 2026, with AICU API. Each participant receives sponsored credits and a personal hackathon key on member.aicu.ai after accepting the terms — text, image, speech and video behind one key, with no payment card. Schools and universities can run the same programme for their groups: a way to teach development in the age of AI.
Participant guide ↗Digital Hollywood University uses AICU API in its “Introduction to Creative AI” e-learning
Digital Hollywood University uses AICU API in its e-learning course “Introduction to Creative AI”, with keys issued for each cohort. AICU also issues certificates of completion for Digital Hollywood University courses, which anyone can verify at cert.aicu.ai.
Verify a certificate ↗Made with Wan 3.0: the music video “Moratorium in Synthetic Real”
AICU made the music video “Moratorium in Synthetic Real” (Saki Noire & AiCuty) with Alibaba’s Wan 3.0 on
Watch on YouTube ↗POST /v1/video/generate, and a Japanese version with improved prompts. Wan 3.0 makes video with sound, 2 to 10 seconds at 480p to 1080p, and a character reference image keeps the face consistent.Qwen-Image 3.0 is available: posters and titles with clean Japanese, English and Chinese lettering
Alibaba’s
Models and prices ↗qwen-image-3.0andqwen-image-3.0-proare onPOST /v1/images/generations. Their strength is lettering: put the words in quotes and say where they go, in which style and on what surface, and they come out as written — a poster with “芸術の秋” brushed across the centre took about 6 seconds. Sizes are 1024x1024, 1536x1024 and 1024x1536.qwen-image-3.0is 1,000 AP an image;qwen-image-3.0-prois 1,300 AP square and 2,350 AP wide or tall. They make images from text only: a request with a reference image or a mask is refused with 400 and not charged (usegpt-image-2.5for references). Prices are inGET /v1/models.AICU Create H3 video: from November 8, 3,000 AP for 3 seconds and 5,000 AP for 5 seconds
From November 8, 2026 (JST), a video made with MiniMax H3 on AICU Create (https://aicu.ai/create/h3/) costs 3,000 AP for a 3-second video and 5,000 AP for a 5-second video. Videos that fail are not charged. The limit of 3 videos per person per day does not change. Until then, H3 videos stay free to try.
AICU Create H3 ↗GPT-6.1 Sol is on the gateway (`gpt-6.1-sol`), plus a `fast` service tier for OpenAI models
OpenAI released GPT-6.1 Sol at DevDay on 2026-09-29: near-Astra results on coding, computer use and document work at one-fifth of Astra's upstream price. It is live here as `gpt-6.1-sol` — 75 AP in / 300 AP out per 1K tokens, 128K max output, reads images, reasoning on (
GET /v1/models (live prices and flags) ↗reasoning_effortlow/medium/high/xhigh; defaultlow). Two upstream limits you should know: function tools are not accepted for the gpt-6 family on `/v1/chat/completions` (we returntools_unsupported_for_modelinstead of forwarding a doomed call; usegpt-5.6-sol/gpt-5.6-terrafor tool calling), andtemperaturemust stay at its default. New on every OpenAI-hosted model:"service_tier": "fast"asks OpenAI for its Fast tier at 2× the AP rate; the response headerX-AICU-Service-Tiertells you what you actually got, and if OpenAI falls back to standard you pay the standard rate.ultrafastis a Responses API tier upstream and is not available through Chat Completions yet.Terms of Service and Privacy Policy version 2026-10b — effective October 22, 2026
We have updated the Terms of Service and the Privacy Policy. The new version takes effect on October 22, 2026; until then, version 2026-10 applies. What changes: A purpose and a simple rule for commercial use. Commercial use is permitted on an account we can reach by e-mail and that has a lawful means of payment, such as a card. A personal account is for a person's research, development, creative work and experiments. Free credits are for evaluation. Free AP is not for resale or business use, and output generated with free credits may include a short notice or advertisement from AICU. Output generated with purchased AP does not. API keys. Use made with your key is your use; do not sell, lend or publish keys. You confirm this when you create a key. Accounts used together. Accounts that share a domain, or that are used by the same organization, may be asked to move to a business or team account. Stored requests. To improve the Service and to detect abuse such as prohibited content and deepfakes, we keep the content of requests for up to 90 days. It is not used to train models. Your output. Rights in output generated from materials you may use are yours. Output you explicitly make public may be published and used by AICU. If you continue to use the Service after October 22, you accept the new version.
Terms of Service ↗Speech on free credits now needs a card on file
To use speech synthesis (
Add a card ↗/v1/audio/speech,/v1/tts/generate) while your account has only free credits, add a card in the dashboard. Adding a card does not charge anything; you are charged only when you buy AP or turn on auto top-up. Accounts that have bought AP, have a card on file, or use a business account are not affected. Text, images and video are not affected. Without a card, speech returns 402payment_method_requiredwith the link to add one. Free credits are for one person to evaluate the Service (Terms 5.3); speech was being used at scale through many new accounts, which we cannot sustain.Video (Gemini Omni): 360p and 720p prices change on October 28
From October 28, 2026 05:00 UTC,
Models and prices ↗gemini-omni-1.1-flashonPOST /v1/video/generateis 450 AP / second at 360p (from 400) and 1,400 AP / second at 720p (from 1,300). 1080p stays at 2,000 AP / second. Omni decides the length of each video itself (about 5 to 10 seconds): the API reserves 10 seconds when the job starts and settles to the length actually produced, returning the difference. Other video models are unchanged. Current prices are inGET /v1/models.Video: Wan 3.0 and Wan 3.0 Prime are available — choose the model on /v1/video/generate
Try it at aicu.ai/create/wan ↗POST /v1/video/generatenow takes amodel. Alongside Gemini Omni (gemini-omni-1.1-flash, the default — the value choice at 720p), you can now use Wan 3.0 (wan3.0-video) and Wan 3.0 Prime (wan3.0-video-prime, about half the time or less) from Alibaba for quality. Both generate sound with the video and keep a character from a single reference image. 480p / 720p / 1080p, 2 to 10 seconds. Prices per second by resolution are inGET /v1/models(typevideo). AP is reserved when the job starts and returned in full if it fails. Try it with AiCuty characters at aicu.ai/create/wan — samples of both models are on the page.Gemini API billing is moving to prepaid for accounts Google has notified — Gemini and Nano Banana keep working through api.aicu.ai
Google AI Studio is moving Gemini API billing from postpaid to prepaid. For the accounts that received Google's notice, the deadline is October 12, 2026: once the prepaid balance is zero, paid Gemini API calls stop, even with a correct key. (If you only use the free tier, Google says there is nothing to do.) If you call the Gemini API with your own key, check the billing page in Google AI Studio. The same models are also on api.aicu.ai, behind one key:
Read the post ↗gemini-flashandgemini-3.6-flashonPOST /v1/chat/completions(OpenAI-compatible — set the base URL and the model name), andnano-bananaandnano-banana-2.1onPOST /v1/images/generations. You pay in AP from one balance that covers every model, and the amount charged comes back with each response. We switched our own Google billing to prepay on October 11 and confirmed image generation afterwards; code that calls api.aicu.ai did not change. The blog post walks through both routes, with samples in Python and curl.The API lineup today: 34 models behind one key, and a round of fixes that make calls say what they do
Models and prices ↗GET /v1/modelsnow lists 34 models: 21 for chat, 6 for images, 4 for speech, 2 for transcription and 1 for video — all behind one key, all billed in AP, each with its price in the same response. The newest is Nano Banana 2.1 for images, added today. What changed around them this month: Transcription is billed by the length the model measured. We used to estimate the length from the text and the file size, which overcharged ordinary speech and undercharged audio with long silences. The price per 10 seconds has not changed. Speech: parameters that would be ignored are now rejected. Sendingvoiceto/v1/tts/generate(which takesslug) used to produce the default voice and a 200; it now returns 400 with the name of the parameter and where to send it instead. The Japanese reading dictionary is public.GET /v1/tts/yomireturns the replacements the server applies before synthesis, so you can check how a name will be read before you generate. The API reference is in English. Every public operation inopenapi.jsonhas an English description. The keys page has samples that run. Speech, text and image, in curl and Python, each verified against production. Images: the file URL needs your key. The response also carries the image inline; the docs now show how to display it in a browser without shipping a key. Prices are inGET /v1/modelsand are reviewed every Wednesday — read them from there rather than keeping a copy.Terms of Service and Privacy Policy updated (version 2026-10) — please confirm in your dashboard
The Terms of Service and the Privacy Policy are no longer drafts. What changed in the Terms: one account per person or organization, and automated systems act under the person or organization that holds the account (3.1, 3.3); business use belongs on a business account — write to api@aicu.ai (3.4); free credits are for personal evaluation (5.3); no resale of the Service without a reseller agreement (4.5); voices need the consent of the speaker (4.6); speech features follow the ElevenLabs terms and prohibited use policy, and we may audit usage for compliance (2.1); what we record, what we store, and when we may look at content (7.1); and a limit of liability (9). The Privacy Policy now describes what the platform does today: the text you send for speech is not kept in usage records, generated output is cached but not archived, and your content is not used to train models. Requests about your data go to api@aicu.ai; abuse reports to api+abuse@aicu.ai. Please confirm: the next time you open the dashboard it asks you to accept version 2026-10 — one click. Your API keys keep working in the meantime. The updated Terms apply from October 11, 2026. Continuing to use the Service after that date means you accept them.
Terms of Service ↗Nano Banana 2.1 is live as `nano-banana-2.1` (14 reference images, 1K/2K/4K, panoramas). `nano-banana` switches to it on 2026-10-21
Google released Nano Banana 2.1 (
Image API docs ↗gemini-nano-banana-2.1, built on Gemini 3.6 Flash) on 2026-10-06. It is on the gateway as `nano-banana-2.1`: up to 14 reference images (4 characters + 10 objects),sizeof1K/2K/4K,aspect_ratioincluding the new panoramas4:1/1:4/8:1/1:8, and better Japanese text rendering.qualitypicks the thinking level (low= minimal,medium,high). Prices measured upstream on 2026-10-07: 1K 1,200 AP (low) / 1,500 AP (medium, high), 2K 1,750 / 1,950, 4K 3,600 / 3,800, plus 60 AP per reference image. Google deprecated Gemini 3.1 Flash Image (the model behindnano-banana) the same day. On 2026-10-21 05:00 UTC the alias `nano-banana` (and `gemini-3.1-flash-image-preview`) will resolve to 2.1; at the default settings that is 1,170 → 1,500 AP per image, hence this advance notice. Pinnano-banana-2.1now, or keep callingnano-bananaand get the new model on that date.New chat models: Qwen3.8 Max and Qwen3.8 Flash are serving now
Two models from Alibaba Cloud Model Studio are on
Models and prices ↗POST /v1/chat/completions:qwen3.8-maxfor long context and harder requests, andqwen3.8-flashfor summaries, classification and anything you run in volume. Change themodeland nothing else — streaming and tool calling work as they do for the other models.qwen3.8-maxis 60 AP per 1K input tokens and 180 AP per 1K output;qwen3.8-flashis 5 and 15. Every price is inGET /v1/models— read it from there rather than keeping a copy. These models can think before they answer, and here that is off unless you ask for it. Upstream it is on by default, the thinking is billed as output, and it does not count againstmax_tokens: we sentmax_tokens: 40and got 3,216 tokens back after 67 seconds. So on api.aicu.ai,max_tokens(ormax_completion_tokens) caps thinking and answer together, and the models answer directly. To have them think, sendreasoning_effort(minimal,low,medium,high); the thinking then comes back inreasoning_contentand is charged as output. Leave room in the cap when you do: a small cap can be spent entirely on thinking, and the answer comes back empty withfinish_reason: "length".The typed-answers endpoint is `POST /v1/jev` (was `/v1/decisions`)
We renamed it during the beta. `POST /v1/jev` replaces
Skill document ↗POST /v1/decisions, and the skill document is at/skills/jev. Nothing else changed — same model (jev-1.13.0), same request body, same price, samellmscope. Why: a *decision* is something a person makes — heavy, discrete, final. What comes back here is a probability: a number per option, an expected level, a 0–1 strength. The old name invited people to read0.92as "settled". It is not settled until you pick a threshold, and picking it is your call. The old path is gone, not deprecated: it had no callers outside AICU, so we cut it rather than leave a second name in place. If you wrote code against it during the beta, change the URL and the responseobjectfield ("decisions"→"jev").JPY prices for AP packs move to the invoice lane rate on Sep 30; USD prices are unchanged
AP packs can be bought from two sellers: AICU Inc. in USD, and AICU Japan K.K. in JPY. The JPY lane exists for customers who need a Japanese qualified invoice (適格請求書), pay by bank transfer against an invoice, or commit an annual budget in advance. From 2026-09-30 05:00 UTC the JPY prices are set from a billing rate of 1 USD = ¥180: Pack S ¥1,700 → ¥1,800 (100,000 AP), Pack M ¥8,400 → ¥9,000 (500,000 AP), Pack L ¥33,500 → ¥36,000 (2,000,000 AP). USD prices do not change: $10.00 / $50.00 / $200.00. Validity stays 24 months. From now on the JPY price is always at least 10% above the USD price converted at the market rate, and we check it every week against the published USD/JPY table. That gap is the price of the invoice lane, not a currency surcharge — if you do not need a qualified invoice, the USD lane is the cheaper one and stays open. Announced nine days ahead, per our price-revision rule (increases get 7 days’ notice and take effect on a Wednesday at 05:00 UTC). The same notice retires two old JPY-only routes that priced the same AP *below* the USD lane: the legacy AP packages (
AP packs ↗POST /v1/stripe/checkout, ¥100 / ¥980 / ¥9,500) and the monthly JPY subscription plans (POST /v1/stripe/subscribe). Both now return HTTP 410 and point at the pack catalogue. No account was subscribed to either.New image models: GPT Image 2.5 Flare and Sunburst are serving now
Image API ↗gpt-image-2.5-flareis the one to reach for by default. It is better than the 2 series and takes about half the time (typically 85 s against 170 s), and it keeps reference-image support, so a character stays the same character across images.gpt-image-2.5-sunbursttrades time for control: finer editing, typically 200 s. Sizes are 1024×1024, 1536×1024 and 1024×1536. Flare at 1024×1024 is 700 AP onlow, 1,600 AP onmediumand 6,400 AP onhigh; the wide and tall sizes are 1,300 / 2,300 / 5,100 AP. Every price is inGET /v1/models— read it from there rather than keeping a copy. Both models regularly run past 100 seconds, which is longer than the edge will hold a connection: if the HTTP call ends in 524, the image is not lost. Keep theidfrom the response (or theX-AICU-Image-Idheader when streaming) and collect it withGET /v1/images/status/{id}. Use the id, notGET /v1/images/recent, whenever more than one generation can be in flight on the same key.Jev: get a typed answer back with a confidence value, not prose
API reference ↗POST /v1/jevsends your state and your questions to Jev (TypeSafe AI) and returns typed answers with a confidence value — a boolean, a number or a choice, in the shape you asked for. No prompt engineering to keep the model from replying in paragraphs, and no parsing of free text at your end. It is the piece you want when a decision sits inside your control flow: should this be escalated, is this record complete, which of these three queues. It uses your existingllmscope, so keys that already work need no change, and it is fast enough to sit on a request path. Billing is per call and small — a two-question decision costs single-digit AP.Beta extended by four weeks: beta keys are issued until Oct 14 and work until Oct 31; production starts Oct 21
We are moving the beta calendar published on Sep 9 back by four weeks. Nothing changes for you before Oct 14. The revised dates, each at 09:00 JST: Sep 23 — Generative AI Expo Vol.6 (Hamamatsucho). Public beta demo at the AICU booth; keys are not switched off. Oct 7 — AP pack purchase opens in USD, and we email every account holder the production guide (commercial declaration, auto-recharge, card registration). Oct 14 — no new beta keys are issued; keys issued from here are production
Issue a key ↗aicu_live_keys. Oct 21 — production operation begins. New keys require a registered card. Oct 31 — beta keys stop working (the same day the beta and the Kumamoto relief programme end). Why: the paid purchase path is not yet open, and the week of Sep 19–24 is taken by the book launch, the expo and the Award first-round deadline. Starting production without a way to pay would be a label, not a launch. Existingaicu_alpha_andaicu_beta_keys keep working until Oct 31; expiry never deletes anything — usage history and balance stay, the key simply returns 401. We email the owner a week before any key lapses. Prices are unchanged by this notice.X (Twitter) gateway: from Sep 23, profile lookups and post reads are 100 AP per upstream fetch (post reads now include attached media)
The X gateway (
X gateway skill →GET /v1/x/*, scopex) lets you read X posts, profiles, timelines and search results through your AICU key. From 2026-09-23 05:00 UTC:GET /v1/x/user/:usernameis billed 100 AP (about $0.01) per upstream lookup, andGET /v1/x/status/:idis also 100 AP (about $0.01) per upstream fetch. Post reads now return the text, author,created_atand the attached images/videos (media[]with URLs and sizes), so a quote card can be built from one call without screenshotting the page — that is why the earlier 10 AP figure (announced 2026-09-13) is revised to 100 AP. Both are cached (profiles 12 hours, posts 10 minutes) and cache hits stay free; failed upstream fetches are not billed. Account-ownership verification stays free.timelineandsearchare unchanged at 2,250 AP per successful call. Revised notice published 2026-09-14, 9 days ahead, per our price-revision rule (increases get 7 days’ notice; only HTTP 200 is billed). Agent-readable spec: https://api.aicu.ai/skills/xGET /v1/models now lists every model we serve — chat, image, speech, transcription — with prices, docs and update dates
Try GET /v1/models ↗GET /v1/modelsis now the single source of truth for everything on the platform. Until today it listed LLMs and images only, andGET /v1/chat/modelswas the only place with prices. Every entry now carries the same fields:type(chat / image / tts / stt), the endpoint to call,stage,pricingwith an explicit unit (per 1K tokens, per image by size and quality, per 1,000 characters, per 10 seconds),docs_url,pricing_urlandupdated_at. The list envelope carriesdocs,pricing_urland the latestupdated_at.GET /v1/models/:idreturns one model, OpenAI-style, and accepts image aliases such asflare.GET /v1/chat/modelskeeps working with the same fields as before; it is now atype=chatfilter of the same array. One label change:stageno longer saysalpha— the alpha programme ended on Sep 9, so models arebetauntil production starts on Sep 23. Also from today: when an account runs out of AP, the owner gets one email per day explaining the pause, sent in Japanese from our Japan reseller (info@aicu.jp) for accounts in Japan, and in English from assistant@aicu.ai everywhere else. Nothing is charged while paused.Key programme: alpha ends Sep 9, beta keys run to Sep 30
Revised on 2026-09-14 — the four dates below were moved back by four weeks (issuance stop Oct 14, production Oct 21, beta keys expire Oct 31). See the Sep 14 notice. The free alpha evaluation closes and metered usage begins on 2026-09-09 (09:00 JST). Four dates follow, each at 09:00 JST inside the Wednesday maintenance window: Sep 9 — keys issued from here carry the
Issue a key ↗aicu_beta_prefix; existingaicu_alpha_keys keep working. Sep 16 — no new beta keys are issued. Sep 23 — production operation begins. Sep 30 — beta keys stop working. The week between Sep 23 and Sep 30 is deliberate overlap: production starts at an event, and we are not switching everyone off on the morning we first run it in public. Move to a production key during that week and keep both live until the new one is confirmed. Once you are on card billing, the next key you issue is a productionaicu_live_key and is not bound to these dates. Expiry never deletes anything — usage history and balance are untouched; the key simply returns 401. We now email the owner a week before any key lapses.Fixed: four models failed unless you set reasoning_effort yourself
Calls to
Model list ↗gpt-5.6-terra,gpt-5.6-luna,gpt-5.4-miniandgpt-5.4-nanoreturned 400 — “reasoning_effort does not support ‘minimal’ with this model” — whenever the request leftreasoning_effortunset. Setting it explicitly always worked, so the models themselves were fine; the default we supply was not. That default exists so reasoning cannot silently consume your wholemax_tokens, and it was sendingminimal, a value the newer dotted generations (gpt-5.4, gpt-5.6) renamed tonone. The default is now chosen by generation instead of by model name, so the next derivative is covered before it ships.gpt-5.6-solwas unaffected and keeps itslowdefault; the cheaper, faster models now default tonone. Passingreasoning_effortyourself still overrides, as before.Provider errors now keep their `param` and `code`
When an upstream provider rejects a request, we passed its message through but replaced
typeandcodewith our own values and droppedparamentirely — so a 400 told you something was wrong without telling you which field. Provider errors are now returned in OpenAI’s own shape,paramandcodeincluded, withprovider_statusadded alongside so you can still tell our layer from theirs. Nothing about status codes orRetry-Afterchanges.Image input confirmed on the gpt-5.6 family — and what tool-calling clients need to know
You can pass images to
OpenCode / Cline setup ↗gpt-5.6-sol,gpt-5.6-terraandgpt-5.6-lunathroughPOST /v1/chat/completionsusing the standard OpenAIimage_urlcontent part, verified end to end against production. Two things to know if you drive these from a coding tool such as Cline or OpenCode. First, tool calling and reasoning cannot be combined on this generation, so the gateway sendsreasoning_effort: “none”whenever your request carriestools— you do not need to set it yourself. Second, reasoning tokens come out ofmax_tokens: with the default effort a short image question spent 74 of 89 completion tokens on reasoning, and a request capped at 40 tokens failed outright. Give image requests room, or sendreasoning_effort: “none”. Note thatGET /v1/chat/modelsstill describes capabilities only in prose, so tools that look for a machine-readable field will assume no image support until you enable it in their config.Video generation arrives Sep 2 — Gemini Omni 1.1 Flash, from text, a still, or two keyframes
From 2026-09-02 you can generate video: text-to-video, image-to-video, first-and-last-frame interpolation, and reference-to-video that holds a character across the shot. Ten seconds per call at 360p, 720p (default) or 1080p, returned synchronously — we measured 35.8s on average for 720p, so there is no job queue to poll. A seed is supported, and the same seed returns the same shot (PSNR 41.3 dB across our runs). Two things to know before you build on it: if you write that someone is speaking but do not supply the line, the model invents dialogue — specify the line, or ask for silence; and 4K is not in this release, because a 4K clip takes about 210 seconds to generate and will not return inside a single request. Free like everything else through Sep 8; billed from Sep 9.
Measured benchmark →Transcription moves to a new backend — more accurate, and it stops rate-limiting you
The default backend for
Japanese STT benchmark →POST /v1/audio/transcriptionsis now OpenAIgpt-4o-mini-transcribe, replacing the previous provider. Two reasons: the old backend returned429under any sustained use — we could not complete twelve retries over fourteen minutes on our own benchmark run — and it placed last in our Japanese STT benchmark at 64.97% CER, because roughly 40% of the body text was dropped at the seams where long audio is split internally. The new default measured 23.40% CER on the same audio, and it costs us slightly less. One behavioural change: the per-request size limit drops from 30 MB to 25 MB, which is the new backend’s own limit; the 30-minute duration limit is unchanged. The previous backend stays available withprovider=sakura, andmodel=whisper-large-v3-turboroutes there too. Prices are unchanged by this switch.Text-to-speech moves to per-character pricing on Sep 9 — 200 AP per 1,000 characters
Speech synthesis is billed per character from 2026-09-09 05:00 UTC, matching how our upstream (ElevenLabs) meters it: 200 AP per 1,000 characters, about $0.02 — roughly one-eighth of ElevenLabs’ own list price for the same voices, and with the character licensing already cleared. Until now we billed on estimated audio seconds, and the estimate ran about 2x the audio we actually produced; that is fixed in the same release. Announced 13 days ahead per our price-revision rule (increases get 7-day notice; cuts are immediate).
TTS pricing →Free alpha evaluation extended to Sep 8 — paid usage starts Sep 9
Everything stays free through 2026-09-08, one week longer than previously planned: we are holding rates flat across the Sep 1–4 student-lecture week. From 2026-09-09 05:00 UTC, usage is metered against your AP balance. Nothing is charged automatically — the platform is prepaid, so if your balance runs out the API simply returns 402, and no invoice or card charge follows. Free tiers, the signup bonus and the first-use bonus all continue to work after that date.
What "billing starts" means →Pricing notice: transcription 10 s = 3 AP from Sep 9 (05:00 UTC)
Speech-to-text (
Japanese STT benchmark →POST /v1/audio/transcriptions) moves from 1 AP to 3 AP per 10 seconds of audio on 2026-09-09 05:00 UTC — about $0.11 per hour of audio, following our cost×3 policy (the current rate has been below provider cost). Update (Aug 26): effective date moved from Sep 2 to Sep 9 — we are keeping current rates through the Sep 1–4 student-lecture week. Announced well past our 7-day-notice rule for increases. Timestamps (verbose_json),srt/vttoutput and the new/skills/sttspec shipped this week; a premium high-accuracy lane (ElevenLabs Scribe) is being evaluated — see the benchmark on the docs blog.Fixed: non-browser clients (e.g. Python-urllib) blocked with 403 on api.aicu.ai
Cloudflare’s Browser Integrity Check was rejecting requests whose
User-Agentlooked non-browser-like — most visiblyPython-urllib/*, the default UA of Python’s standard-library HTTP client — with403(error code: 1010), even for public endpoints like/healthand/openapi.json. This is fixed forapi.aicu.ai; if you were setting a fakeUser-Agentto work around it, that’s no longer necessary. (We kept this protection in place on our other, non-API domains.)Fixed: GET /v1/usage was missing TTS and image requests
The self-serve usage summary (
View your usage →GET /v1/usage) was silently scoped to LLM calls only, so text-to-speech and image-generation requests didn’t show up insummary.requestsorby_model— even though they were billed correctly and appeared inGET /v1/usage/logsand the dashboard. Fixed: the summary now covers every billed service.ElevenLabs v3 — emotional TTS, plus a dedicated compat endpoint
Audio API docs →POST /v1/audio/speechand the newPOST /v1/el/ttsnow acceptmodel: "eleven_v3"for ElevenLabs’ newest model — inline audio tags like[excited]or[whispers]shape delivery (5,000-char limit per request, vs 10,000 for the defaulteleven_multilingual_v2). Unrecognizedmodelvalues now return a clear 400 instead of silently generating with the wrong model. Also fixed:POST /v1/tts/generate(theslug-based legacy path, e.g.slug: "luc4") had been pointing at a retired backend and always failing — it now runs on the same ElevenLabs gateway as everything else. Verified against production with real audio for both models.Gemini pricing update
The default
See pricing tiers →gemini-flashnow maps to Gemini 2.5 Flash-Lite at 3 / 12 AP per 1K tokens (input / output) — lighter and cheaper for high-volume work. A newgemini-3.6-flashis added at 45 / 225 AP for top-quality reasoning. All rates follow our cost×3 policy.Kimi-K2.7-Code (Powered by Sakura AI Engine)
A coding-specialized model —
OpenCode setup →kimi-k2.7-code. OpenAI-compatible and verified working in OpenCode (v1.18.9) through api.aicu.ai. Also works from Cline and Continue. Free during public preview; domestic inference, no training on your data.API keys required · reasoning models improved
All
Get an API key →/v1endpoints now require a Bearer API key (issue one in the dashboard; public listing endpoints stay open). Reasoning models (gpt-5*) now honorreasoning_effort(minimal/low/medium/high) and default tominimalwhen unset, so short prompts return content instead of spending the budget on reasoning tokens.Reiwa 8 Kumamoto Earthquake — free API for relief apps
Free access to our AI API infrastructure (OpenAI-compatible LLMs, multilingual text-to-speech, Whisper speech-to-text, image generation) for developers building disaster-support apps, through 2026-10-31.
AICU Call ↗Audio API — multilingual TTS + Whisper STT
Text-to-speech via ElevenLabs voices (elena / mei / mina / nao / saki) at
Audio API docs →POST /v1/audio/speech, and speech-to-text via Sakurawhisper-large-v3-turboatPOST /v1/audio/transcriptions. Both are OpenAI-SDK compatible.Image API with C2PA provenance
Generate images at
Image API docs →POST /v1/images/generations(GPT-Image-2 and Nano Banana 2), and detect embedded C2PA content-credentials atPOST /v1/images/inspect.