Venice.ai API

Run text inference based on the supplied parameters. Supports multimodal inputs including text, images (image_url), audio (input_audio), video (video_url), and files (file) for compatible models. File inputs (PDF, DOCX, PPTX, XLSX, TXT, etc.) are automatically extracted to text before being sent to the model. Long running requests should use the streaming API by setting stream=true in your request. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

**Beta**: This endpoint is currently in beta and may be unstable. Request/response schemas and behavior may change without notice. It is available to all API users. Create a model response in OpenAI's Responses API format. Works with every Venice text model and returns typed output items such as `reasoning`, `message`, and `function_call`. Set `stream: true` to receive Server-Sent Events. Most OpenAI models are forwarded to OpenAI's Responses API natively. Other models, and requests that use Venice features such as web search or characters, are translated through Chat Completions, where some Responses features are not supported. Send `x-venice-responses-mode: native` or `compat` to choose the mode, and read the `x-venice-responses-lane` response header to see which one served the request. Venice does not store responses, so send the full conversation in `input` on every request. E2EE-capable models are not supported on this endpoint; use `/api/v1/chat/completions` with the required E2EE headers instead. See the [Responses API guide](https://docs.venice.ai/guides/features/responses-api) for the features and limitations of each mode. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Generate an image based on input parameters **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Generate an image based on input parameters using an OpenAI compatible endpoint. This endpoint does not support the full feature set of the Venice Image Generation endpoint, but is compatible with the existing OpenAI endpoint. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Upscale an image to a higher resolution. The image can be provided either as a multipart form-data file upload or as a base64-encoded string in a JSON request. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Edit or modify an image based on the supplied prompt. The image can be provided either as a multipart form-data file upload or as a base64-encoded string in a JSON request. For models with resolution tiers that require explicit dimensions, omit aspect_ratio or set it to auto to infer the closest supported aspect ratio from the input image; provide aspect_ratio directly when exact output dimensions are required. Use output_format to request jpeg, jpg, png, or webp output. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Edit or modify an image using multiple layered inputs (base image plus masks/overlays), up to the selected model's maximum input images (see the capabilities.maxInputImages field in GET /api/v1/models). **Supported input formats by Content-Type:** - **multipart/form-data**: Only file uploads are supported. Send images as form file fields. - **application/json**: Base64 strings and URLs are supported: - Raw base64 string: `"iVBORw0KGgoAAAANSUhEUgAA..."` - Data URL: `"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..."` - HTTP/HTTPS URL: `"https://example.com/image.png"` For models with resolution tiers that require explicit dimensions, omit `aspect_ratio` or set it to `auto` to infer the closest supported aspect ratio from the first input image. Provide `aspect_ratio` directly when exact output dimensions are required. Use `output_format` to request jpeg, jpg, png, or webp output. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Remove the background from an image using AI. The image can be provided either as a multipart form-data file upload, as a base64-encoded string in a JSON request, or as an image URL. Returns a PNG image with transparent background. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Create embeddings for the supplied input. `input` must be a string or an array of strings. Token arrays (integer IDs) are rejected with HTTP 400. LangChain `OpenAIEmbeddings` sends token arrays unless `check_embedding_ctx_length` is `False`. Wrappers that construct `OpenAIEmbeddings` without that flag, including gpt-researcher's `openai` provider, still send token arrays. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

**Beta**: This endpoint is currently in beta and may be unstable. Request/response schemas and behavior may change without notice. Evaluate a `state` against a map of typed `questions` and get back structured `answers`, one per question. Question types: `noul` (yes/no probability), `choice` (pick one option with full probability distribution), and `score` (probability-weighted position on an ordered rubric). Every question is evaluated in parallel against the same state. This endpoint returns typed judgments, not generated text. It is also available at `/api/v1/systemone` for drop-in compatibility with TypeSafe SDKs (set `TYPESAFE_BASE_URL=https://api.venice.ai/api`). **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

**Beta**: Alias of `POST /decisions` mirroring the upstream TypeSafe path, so TypeSafe SDKs work with a base URL swap (`TYPESAFE_BASE_URL=https://api.venice.ai/api`).

Converts text to speech using various voice models and formats. **Output formats:** Format support is model-specific. Read `model_spec.supported_formats` and `model_spec.default_format` from GET /models?type=tts. Omitting `response_format` uses the model default; requesting an unsupported format returns HTTP 400. **Voice cloning:** Some TTS models (e.g. `tts-chatterbox-hd`) accept a cloned-voice handle in the `voice` field. Mint a handle by uploading a reference audio sample to POST /v1/audio/voices, then pass the returned `vv_<id>` value as `voice` here paired with the same `model`. Supported voice-cloning models advertise the `voice_cloning` capability on GET /models?type=tts. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Transcribes audio into the input language. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Clone a voice from an audio sample. Returns a voice handle (`vv_...`) that can be passed to POST /v1/audio/speech as the `voice` parameter to synthesize speech in the cloned voice; the handle must be paired with the same `model` used to create it. Supported models: - `tts-chatterbox-hd` (Resemble AI) — accepts MP3, WAV, FLAC, and M4A samples. Zero-shot: no persistent voice template is derived; the reference audio is uploaded to a TTL’d storage object and re-read on each synthesis call. Handles expire automatically after 7 days. Available to regular users. - `tts-minimax-speech-02-hd` (MiniMax) — accepts MP3 and WAV only. Persistent: the upstream provider derives a custom voice template that survives across synthesis calls. Auto-deleted after 7 days without use; each successful TTS request resets the window. Currently limited access — contact support@venice.ai if you need it enabled. Samples in containers outside the per-model allowlist are rejected with HTTP 400 before any upload. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Delete a video generation request from storage after it has been successfully downloaded. Videos can be automatically deleted after retrieval by setting the `delete_media_on_completion` flag to true when calling the retrieve API. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Queue a new video generation request. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Retrieve a video generation result. Returns the video file if completed, or a status if the request is still processing. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Extracts text from a document file. Supports PDF, DOCX, PPTX, XLSX, and plain text formats. Upload a file via multipart/form-data. **Privacy:** Text parsing runs entirely in-memory on Venice's infrastructure with zero data retention. Documents are processed and immediately discarded — no content is stored or logged. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Mark an audio generation request as complete and clean up the generated media from storage. Call this after you have successfully downloaded the audio if you did not set delete_media_on_completion in the retrieve request. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Queue a new audio generation request. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Retrieve the status or result of an audio generation request. If the audio is still being generated, returns processing status with estimated time. If complete, returns the audio data. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Queue a speech-to-speech conversion, re-recording a source recording in a different voice while preserving its delivery, pacing and timing. Supply the source recording exactly one of two ways: as a `multipart/form-data` upload in the `file` field, or as an http(s) `audio_url` in a JSON body. Supplying both, or neither, is rejected. When you pass a URL, Venice fetches and validates the bytes itself and forwards only those bytes to the provider — the URL is never handed onward. The recording is validated from its binary signature rather than its filename or declared content type, so the accepted containers published as `accepted_audio_formats` in `/models` are enforced on the actual bytes. Recordings above the size limit or longer than `max_source_audio_duration_seconds` are rejected before any charge. Billing is based on the length of the source recording, measured server-side and rounded up to the next whole minute; the measured length is returned as `duration_seconds` so it can be reconciled against `/audio/voice-changer/quote`. A queued conversion has already been charged, so this request is not safe to retry: if the response is lost, poll `/audio/voice-changer/retrieve` rather than queueing again. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Poll a queued voice conversion. While the conversion is still running this returns `status: "PROCESSING"` as JSON; once it finishes it returns the converted audio bytes. A completed response carries the audio alongside `x-venice-audio-format`, `x-venice-audio-duration`, `x-venice-inference-time`, `x-venice-model-id` and `x-venice-model-name`. Only the account that queued a conversion can poll it. Any other `queue_id` — unknown, belonging to someone else, malformed, or older than the retrieve window — is reported the same way, as an invalid request id. If the provider fails the conversion, the charge is refunded automatically and the response reports whether that refund succeeded in `credits_refunded`. The outcome is recorded, so polling again replays the same result rather than refunding twice. When `credits_refunded` is `false` and the conversion was charged, the `error` message says what to do about it: a refund that was attempted and did not go through asks you to contact support and is retried by polling again, and a request id the provider no longer recognises is reported as too weak a verdict to reverse a charge on. A conversion that was never charged, or one already delivered, reports `false` with nothing further. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Release the provider-held media for a finished conversion. Call this once the converted audio has been retrieved. Safe to call more than once, and safe to call for a conversion that never reached the provider. Passing `delete_media_on_completion: true` to `/audio/voice-changer/retrieve` does the same cleanup inline and makes this call unnecessary. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Proxy a JSON-RPC request to a supported blockchain node and bill per credit. ## Request shapes - **Single request**: a JSON-RPC 2.0 object (`{ "jsonrpc":"2.0", "method":"…", "params":[…], "id":1 }`). - **Batch**: an array of up to 100 JSON-RPC 2.0 objects. If any item references an unsupported method, the entire batch is rejected with 400 and the offending methods are listed. ## Supported methods Methods are classified into three pricing tiers: - **Standard (1×)**: `eth_call`, `eth_getBalance`, `eth_blockNumber`, `eth_sendRawTransaction`, `eth_getLogs`, `net_version`, `web3_clientVersion`, ERC-4337 bundler methods (`eth_sendUserOperation`, etc.), chain-family extensions (`zks_*`, `linea_*`, `bor_*`, `starknet_*`), and Solana methods (`getAccountInfo`, `getBalance`, `getLatestBlockhash`, `sendTransaction`, etc.). - **Advanced (2×)**: `trace_*`, `debug_*`, `txpool_inspect`, `txpool_status`, `arbtrace_*`. - **Large (4×)**: `trace_replayBlockTransactions`, `trace_replayTransaction`, `txpool_content`, `arbtrace_replay*`, and Solana `getLargestAccounts`, `getSupply`. Stateful filter methods (`eth_newFilter`, `eth_getFilterChanges`, `eth_uninstallFilter`, etc.) are **not supported** — they break on a load-balanced HTTP proxy because filter state is pinned to a single upstream backend. Use `eth_getLogs` instead. WebSocket-only methods (EVM `eth_subscribe`/`eth_unsubscribe` and Solana `*Subscribe`/`*Unsubscribe`) return 400 because this proxy is HTTP-only. ## Pricing Credits consumed per call = `baseCredits[chain] × methodTier`. `baseCredits` is 20 for most EVM chains (Ethereum, Base, Optimism, Arbitrum, Polygon, Linea, Avalanche, BSC, Blast) and Starknet; 30 for zkSync Era and Solana. The USD price per credit is `~7e-7` — a single standard EVM call costs ≈ $0.000014 and a large trace-replay costs ≈ $0.000056. Per-request errors at the JSON-RPC layer (HTTP 200 with an `error` field in a response item) are billed at 5 credits instead of the full method tier — a small concession for methods not supported on a given chain or bad-parameter responses. ## Rate limits Two caps apply per caller: - **Requests per minute**: 100 on the paid tier. - **Credits per rolling 24 hours**: 10,000,000 on the paid tier. When either cap is exceeded, the request returns 429 with a `customMessage` identifying which cap tripped. The per-minute cap also sets the `X-RateLimit-*` response headers. ## Idempotency Set the `Idempotency-Key` request header to any string matching `[A-Za-z0-9_-]{1,255}` to enable safe retries. The response is cached for 24 hours keyed on `(user, idempotency-key)`; replaying the same key with the same body returns the cached response with `Idempotent-Replayed: true`. Reusing the same key with a different body returns 400 to prevent silent corruption. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Top up your Venice credit balance using a `PAYMENT-SIGNATURE` header (the legacy `X-402-Payment` and `X-PAYMENT` header names are also accepted). If the header is missing, the endpoint returns payment requirements. This is the primary x402 payment endpoint. It currently returns Base and Solana USDC payment options in the `accepts` array. All inference endpoints (chat, image, audio, video) consume from the credit balance you establish here.

Scrape a web page and return its content as markdown. Supports most public web pages; some sites (e.g. X/Twitter, Reddit) that block automated access are rejected immediately. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.

Search the web and return results directly. Returns structured search results including titles, URLs, content snippets, and dates. **Search providers:** - `brave` (default) — Brave Search with Zero Data Retention (ZDR). Search queries are never stored or logged by the search provider. - `google` — Google Search with anonymized queries. Searches are proxied through Venice's infrastructure so that your identity is not associated with the search request sent to Google. Venice does not store or log search queries. **Authentication:** This endpoint accepts either a Bearer API key or a `SIGN-IN-WITH-X` header for x402 wallet-based authentication. The legacy `X-Sign-In-With-X` header is also accepted during migration. When using x402, a `402 Payment Required` response indicates insufficient balance and includes top-up instructions.