Octave 2
An LLM-based TTS model that interprets the meaning of the text to choose pronunciation, pitch and tempo rather than reading flatly.
- Weights
- Not released
- API
- Yes
- Commercial use
- Permitted subject to conditions
It supports designing a voice from a text description, cloning from as little as 15 seconds of audio, and phoneme/word-level timestamps. Currently in preview alongside the established Octave 1.
Voice cloning is documented as possible from ~15 seconds of audio; the pages fetched did not state consent verification requirements or explicit commercial-use terms for cloned voices.
- Requires consent from the person depicted. The voice-cloning documentation page returned 404 at time of check, so consent policy is unverified
Specification
- What it takes in
- text, audio
- What it makes
- audio
- Controls
- emotion-control, style-prompting, voice-cloning, timestamps, streaming-low-latency
- Weights
- Not released
- API
- Yes
- Pricing
- Per 1,000 characters, decreasing by tier: $0.15 (Free/Starter), $0.12 (Creator), $0.10 (Pro), $0.05 (Scale). Plans: Free $0, Starter $3, Creator $7/$14, Pro $70, Scale $200, Business $500, Enterprise custom. Included characters range 10,000 (Free) to 10,000,000 (Business). EVI speech-to-speech billed per minute from $0.07 down to $0.04.
- Released or updated
- not established
- Languages
- 11, not individually listed by the vendor
Strengths
- ~100ms model latency on Octave 2 (vs ~200ms on Octave 1)
- Voice design from a text description, no reference audio needed
- Long-form context preservation for podcast and VO work
- Phoneme and word-level timestamps
Weaknesses
- Octave 2 is in preview
- 11 languages, and voice design is English-only
- 5,000-character cap per utterance, max 5 generations per request
- Consent and commercial-use terms for clones not documented on reachable pages
- Base rate ($0.15/1K chars) is higher than most competitors before volume tiers
Best for
- emotionally directed narration
- character work driven by script context
- podcast and long-form VO
Where the sources disagree
Recorded rather than resolved. Picking the tidier number would hide the fact that the vendor's own pages do not agree.
- Hume's docs describe Octave as 'the first text-to-speech system built on LLM intelligence' - a marketing claim, not recorded as a spec.
What to understand before using it
Explainers naming the mechanisms this model's own documentation mentions.
See it from the vendor
Shotcraft hosts no sample output. Judging a model means watching its own showreel with your own eyes, and a clip re-hosted here would be a clip we chose.
Nearby in the directory
Azure AI Speech text-to-speech (Neural, Neural HD, MAI voices)
Microsoft's enterprise speech synthesis service, with 200+ prebuilt neural voices across 80+ locales plus tiers for HD voices, multilingual voices, mu
Voice and speech
Sonic 3.5
Cartesia's current TTS model, documented at sub-90ms first-byte latency across 42 languages. It is tuned for conversational delivery and for reading a
Voice and speech
XTTS-v2
An open-weights multilingual TTS and voice cloning model covering 17 languages at 24kHz, still widely used for local voice cloning. Coqui Inc. shut do
Voice and speech