Whisper.Api vs Voyage AI

Voyage AI is a retrieval layer, not a transcription tool. Its lineup splits three ways: general-purpose embedding models (voyage-3.5, voyage-3.5 lite), domain-specific models tuned for finance, legal, and code, and company-specific models fine-tuned on proprietary data and internal lingo. The Voyage 4 series and voyage-multimodal-3.5 extend retrieval to images and text in one pipeline, while voyage-context-3 keeps chunk-level detail with global document context. Rerank-2.5 and rerank-2.5-lite add instruction following, and a Batch API handles large-scale embedding workloads. Engineering choices target cost and scale: embeddings run 3x-8x shorter in dimensionality, models are 4x smaller, and there's 32K-token context. Whisper.Api is a deployment wrapper around whisper.cpp: a Deepgram-compatible REST API plus real-time WebSocket streaming, audio auto-detection (PCM, WebM, OGG, FLAC), speaker diarization, SRT/VTT subtitle export, user-level API keys via CLI, CPU transcription, Docker deployment, and offline operation with multiple GGML models. The capability sets don't overlap: Voyage answers 'what should this query match,' Whisper.Api answers 'what was said.'

The pricing models are fundamentally different and not directly comparable. Voyage AI lists pricing_type as 'contact' — you talk to sales, and cost scales with usage of embedding and reranking models across general, domain-specific, and fine-tuned tiers, plus batch jobs. The cost story there is inference and vector economics: shorter embeddings cut storage and search bills, and Voyage cites 2x cheaper inference on its 4x smaller models, so the savings show up in your infra as much as in the invoice. Whisper.Api is free to self-host: no license fee, no per-minute transcription charge, and no vendor contract. What you pay instead is in engineering time and hardware — deploying the server, running whisper.cpp on CPU, managing GGML models, scaling, and performance tuning yourself. For a buyer, the honest read is that Voyage is a paid, procurement-driven retrieval service while Whisper.Api trades dollars for operational ownership. Neither one's pricing tier tells you anything about the other's total cost of ownership.