Clipia

Chat with a text LLM and get reply text, token usage and credit cost. Supports strict JSON Schema output and an idempotency key for safe transport retries. Pass either a single prompt or a full messages array. Charged in credits from the connected account.

Stitch 2–20 finished video scenes into ONE final clip server-side (scenes are normalized to a common frame/fps, concatenated, optional soundtrack replaces scene audio). Pass request_id of your COMPLETED video generations (or media.clipia.ai URLs). Returns request_id (cmp_*) — poll with wait_generation until COMPLETED, then output.video.url is the final mp4. Typical full-video flow: generate_scenario → generate_video per scene → compose_video.

Edit an EXISTING presentation instead of regenerating it: change slide text, title, theme, add/remove/reorder slides. The deck is rebuilt from its saved spec and previously generated illustrations are reused without another illustration charge, so a text fix costs only the small render fee. Use it for ANY change request about a deck that already exists; generate_presentation is only for a brand-new deck. All "index" values are 0-based positions in the CURRENT deck. An illustration is redrawn (and billed) ONLY if you change that slide's image.prompt, list it in regenerate_images, or set regenerate_all_images=true. Returns request_id (prs_*) — poll with wait_generation until COMPLETED.

Generate speech from text (text-to-speech): pick a voice and language, get an audio file back. Returns request_id and cost in credits immediately — narration is quick, poll with wait_generation until COMPLETED, then output.audio.url is the file and output.audio.format its format (mp3 or wav, depending on the model). Pass the resulting request_id as voiceover_request_id in compose_video to narrate a multi-scene video, or use the file on its own.

Generate image(s) from a text prompt, optionally with reference images (editing / image-to-image). Waits briefly for completion and usually returns the finished image inline (URL + small preview). Cost in credits is returned.

Generate background music / a soundtrack from a text description: instrumental (no vocals) by default — ideal to play under narration — or a song with your own words (pass lyrics). Returns request_id and cost in credits immediately; music takes ~1–3 min, so poll with wait_generation until COMPLETED, then output.audio.url is the mp3. Pass the resulting request_id as audio_request_id in compose_video to score a multi-scene video, or use the mp3 on its own. Describe mood/genre/tempo/instruments in the prompt (English works best).

Generate a BRAND-NEW slide presentation (editable PPTX + PDF + PNG previews) from a DeckSpec you compose. To change a deck that already exists (slide text, title, theme, add/remove/reorder slides) use edit_presentation instead — it reuses the illustrations already generated instead of billing them again. Returns request_id (prs_*) — poll with wait_generation until COMPLETED, then output has pptx_url, pdf_url, preview_urls[]. Write the full spec: title, language (ru/en), theme, optional brand, and 3–20 slides. ALWAYS pick the theme that fits the SUBJECT of the deck; the deck belongs to the user, so never default to the Clipia brand palette unless the deck is about Clipia itself. Set brand.name/brand.website only if the user names their company — otherwise slides stay unbranded. Each slide has a "layout": cover, section, bullets, two-col, image-full, quote, stats, or closing. For illustrations set image.prompt (English) — generated by our image models and billed as credits; skip images for text-only slides. Costs credits (render fee + illustrations).

Plan a multi-scene video from a brief: an LLM director returns a structured scenario — per-scene English video prompts with durations, plus a soundtrack prompt. Feed each scene prompt to generate_video, then stitch the finished clips with compose_video. Charged in credits like chat (LLM usage).

Start a video generation: text-to-video (prompt only), image-to-video (image_url; add end_image_url for a first+last-frame clip), reference-to-video (image_urls / reference_video_urls on multi-reference models), lip-sync / talking photo (image_url + audio_url on lip-sync models), motion control (image_url + video_url on motion-control models — the driving video sets both the clip length and the cost: the server measures it and prices the call by its seconds, duration has no effect) or video edit / transform (video_url on video-edit models). Pick the model by capability in list_models; model-specific fields go in input (see get_model). The default model is tuned for quality and is not the cheapest one — compare reference prices in list_models. Returns request_id and cost in credits immediately — video renders take 1–10 minutes, poll with wait_generation.

Get the connected account credit balance and 30-day API call count plus net settled ledger spend. API keys cover their operations; OAuth covers account API/MCP operations. Reservations and refunded charges are excluded.

Get the current status/result of a generation without waiting. When COMPLETED, output.images[].url is the inline webp preview and output.images[].original_url is the full-quality PNG/JPG.

Get model details and pricing. Generation models include input_schema.parameters (name/type/options with human-readable labels, defaults and ranges — use them for the input object: they are the parameters the model accepts, and a generate call with a misspelled name or a parameter the model is known not to apply is rejected; some of them change the cost) and reference pricing (TTS per 1000 characters, plus text_limits for the allowed text length); text models include context limits and per-1K input/output credit rates.

List available AI models with type (text/image/video/audio), capabilities and pricing in credits. Text models are priced per 1K input/output tokens; generation models show reference pricing; character-priced TTS declares pricing.unit_size. Use before chat or generation when the user asks for a specific model or wants to compare options or prices. The result is one page: total is the number of models matching the filters, and truncated=true means more models follow — narrow with type or search, or call again with offset=next_offset.

Search 3500+ curated prompt templates (hybrid text+semantic search, Russian or English query). Each result has a ready-to-use prompt and a recommended model — pass them to generate_image/generate_video.

Wait for a generation to finish (long-poll up to wait_seconds, then returns current status). Call repeatedly until status is COMPLETED, FAILED or CANCELED. Returns output URLs (and an inline image preview) when done.