Skip to content
- Video modes:
POST /v1/videos/generations takes a mode: text (from the prompt alone), image (animate your most recent image, with an optional last frame), reference (up to 8 recent images and 3 recent clips as references) and edit (change an earlier clip). New fields mode, explicit, end_frame, image_count, reference_video_count, source_video and video_origin; the response gives the mode that ran. New models wan-edit, sd-fast and sd-edit. Clips run 5 to 15 seconds and are priced per second by model and mode. Changed: without mode, a call with no recent image now makes a clip from the prompt instead of returning image_required. See Video and Video models.
- Voices and the speech model from code: new
GET /v1/audio/voices lists the 200 voices with their gender, age, accent, use cases, style, description and a sample (preview_url), filtered by gender, age, accent, use_case and style. New GET /v1/audio/models describes aderra-voice-1: its 60 languages, formats, speed range, limits, price and how many voices match each filter. Both take any Aderra key and are free. An unknown voice now points to them. See Voices.
- Voice samples: every voice on the Voices page has a sample to play, and the page filters by age, accent, use case and style as well as gender.
- Text to speech docs reorganised into a section: Get started, Voices, Languages, Audio formats, Speed, Streaming and long text, Limits and quotas, Errors and Best practices.
- Text to speech: new
POST /v1/audio/speech, OpenAI-compatible: 200 voices that each speak every supported language, automatic language detection, mp3, wav, pcm, aac (as MP4), opus and flac, a speed setting, and text up to 4,096 characters joined into one file. The audio is streamed as it is generated: the first audio arrives in about half a second. API keys only. See Text to speech.
- API keys run without an Aderra persona. Your system prompt,
system and instructions pass through untouched; Aderra adds one line, “Never produce any sexual content involving minors.”. Terminal keys keep Aderra’s agent instructions. See System prompts.
- Video price:
minimax-h3-open is now $0.10 per second; minimax-h3 and minimax-h3-spicy stay at $0.08. See Video models.
- Every model listed: the Models page now lists every chat, image, video and speech model by id, with its price, sizes and default.
- Chat prices in the docs: the Pricing page lists each chat model’s price per million tokens, DeepSeek V4.1 Flash included.
GET /v1/usage/prices stays the authoritative source.
- Prices from code:
GET /v1/usage/prices now answers any API key on https://api.aderra.ai and lists chat, image, video and speech. Media, like chat, costs the same through the API as on a plan. See Pricing.
- New API documentation: this section, with an API reference and an OpenAPI 3.1 file. The former single page,
/docs/developers/api/, now leads here.
- Your system prompt is kept. Aderra’s agent instructions are now added to the caller’s own system prompt or instructions instead of replacing them. (Since 2026-11-11 this applies to terminal keys only.)
- Small
max_tokens on GLM-5.3 Flash: when you set max_tokens below 16,384 and say nothing about thinking, Aderra sends reasoning_effort: "minimal", so the budget goes to the answer.
- Empty answers are free: a chat reply with no visible text or tool call is not charged.
- One price for API and plans: every chat model costs the same through the API as on a plan.
- Rate limits: every key has a per-minute limit (60 by default, up to 600), and an account can have 8 requests in progress at once across its keys.
- Launch of the pay-as-you-go API: a prepaid API wallet, API keys with monthly caps and rate limits, and per-key usage and request history on the Developers page.
https://api.aderra.ai becomes the API’s address. https://aderra.ai/v1 stays as a permanent alias.- Chat Completions, Messages, Responses, Models, Images and Video available with API keys.
