Orchard vs ElevenLabs.

ElevenLabs meters in characters of input text, which is great when you forget how quickly a podcast script piles up characters. A 5-minute podcast is roughly 5,000 characters; at ElevenLabs Creator ($22 / 100K chars), that's 100 such podcasts a month before you start paying overage at $0.30 per 1K chars (i.e. ~$0.30 per podcast minute). Orchard meters in minutes of output audio, which is what you actually shipped to the user:

On low volume (under ~15 min/mo of synth) ElevenLabs Free is fine. Anything past that, especially anything that needs voice cloning, and Orchard saves 10× or more on the same workload.

ElevenLabs Professional Voice Cloning is the industry leader for single-voice fidelity — a 30-minute reference yields a voice with ~0.92 speaker similarity on most benchmarks. The catch: it locks to the $99/mo Pro tier and a maximum of 10 voices.

Orchard ships two cloning engines depending on the language: F5 for Spanish (with a rioplatense fine-tune we trained on porteño speech — nobody else has this) and XTTS for the other 16 languages. Reference is 6-60 s, and once cloned, every subsequent synth in that voice is on the per-minute plan rate. Voice limits scale by plan: 1 voice on Free, 3 on Hobby, 10 on Starter, 50 on Pro.

• Spanish (rioplatense): Orchard wins on accent fidelity — ElevenLabs uses generic Spanish.
• Spanish (neutral): Roughly comparable quality, Orchard ~5× cheaper per synth.
• English: ElevenLabs Professional is sharper on long-form narration if you can spend $99+ and have the 30 min reference.
• Other 16 languages (XTTS): Orchard covers them with a single API; ElevenLabs has more voices but the per-synth cost makes batch generation expensive.

ElevenLabs covers 32 languages with their Multilingual v2 model. Orchard ships 17 — the gap is real if you're targeting Bulgarian, Vietnamese, or Tagalog. Where Orchard pulls ahead is per-language depth: we ship a fine-tuned rioplatense Spanish (F5-AR) that captures the porteño accent specifically, and our Spanish synthesis quality across all LATAM dialects is meaningfully better than generic-Spanish offerings.

• If you ship in EN / ES / PT / FR / DE / IT / HI → Orchard wins on price by a factor of 10×.
• If you need Bulgarian, Tagalog, Korean or other less common languages → ElevenLabs has broader coverage today.
• If you need a voice that nails Argentine Spanish specifically → only Orchard ships a fine-tune for it.

ElevenLabs is a TTS-first company that recently added Scribe (STT) and Studio (dubbing) as separate products with separate metering. Orchard ships three products on the same balance:

Unlike Whisper (where we're API-compatible with the OpenAI SDK), ElevenLabs uses its own request shape. Migration is a code change, not an env var swap. Three calls map cleanly:

// ElevenLabs
await elevenlabs.textToSpeech.convert(voice_id, {
  text: "Hola, qué tal", model_id: "eleven_multilingual_v2"
});

// Orchard
await fetch("https://api.orchardrun.com/v1/tts/generate", {
  method: "POST",
  headers: { Authorization: `Bearer ${ORCHARD_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({ text: "Hola, qué tal", voice_id: "es_MX-claude" }),
});

For voice cloning, our API takes a 6-60 s reference WAV plus a short text prompt and returns the voice_id you can then synth against. The cloning step is a one-time call per voice; subsequent synths are billed on the normal per-minute rate. We document the full request/response shapes in /docs#clone-voice.

If your product is built around a single, premium-quality cloned voice — audiobook narration, premium agent characters, branded podcast hosts — and the $99/mo Pro tier doesn't move the needle on your budget, ElevenLabs Professional Voice Cloning is hard to beat on raw fidelity. Same for Studio Dubbing, which we don't have at all.

You should also stay on ElevenLabs if you specifically need their 1,000+ pre-made voice marketplace, their Voice Design (text-to-voice from a description), or coverage of one of the 15 languages we don't ship yet (Korean, Vietnamese, Bulgarian, etc.).

For everything else — Spanish-heavy workflows, high-volume generic-voice synthesis, products that also need STT or diarization, teams who prefer per-minute pricing to character metering — Orchard is the economically obvious choice.