deAPI

deAPI lets you generate images, create videos, clone voices, and transcribe YouTube videos — all through one AI inference API and one billing account. It serves 25 open-source AI models across every non-LLM modality, so you can stop juggling five different providers just to ship a single product.

There are two ways to connect. The OpenAI-compatible endpoint is a drop-in replacement for the OpenAI Python and Node SDKs: change the base_url and api_key, keep everything else. The native REST v2 API covers everything the OpenAI spec doesn't, including video, music, OCR, voice cloning, voice design, and more.

For images, deAPI offers 7 models: Flux.1 Schnell, FLUX.2 Klein, Z-Image-Turbo, Z-Anime Distill, plus image editing, background removal, and AI upscaling — starting from $0.00136/image.

For video, 7 models cover text-to-video, image-to-video, and audio-to-video with LTX. Wan2.2-Animate handles character replacement, and video upscaling goes up to 4x — starting from $0.00174/clip.

For audio, 9 models handle three jobs. TTS: Kokoro (40+ voices, 7 languages), Chatterbox (23 languages), and 3× Qwen3 TTS, with voice cloning from a 5-15s clip and voice design from text, from $0.77/1M chars. Music: ACE-Step generates full tracks with lyrics, BPM, key, time signature, and style transfer. Transcription: WhisperLargeV3 supports 80 MB uploads (3x OpenAI) and URL ingest from YouTube, TikTok, Twitch, Kick, and X, from $0.021/hour.

For text, 2 models round things out: OCR with Nanonets, and embeddings with BGE-M3 (1024 dims, 8192 tokens) for RAG and semantic search, from $0.000068/1K tokens.

Pricing is pay-as-you-go with no subscription. You get $5 in free credit on signup, no card required. Any top-up unlocks Premium: 300 RPM per endpoint with no daily cap.

Built for developers: a Python SDK, an n8n node, and an MCP server for Claude, Cursor, and ChatGPT. Get results via HMAC-signed webhooks, WebSocket live previews, or polling. A distributed verified worker network keeps your data in RAM only — no disk writes.