Audio & voice · pricing
Boson AI suits developers building real-time voice agents, text-to-speech or talking avatars who prefer pay-per-use API pricing and an OpenAI Realtime-compatible interface.
Boson AI sells its Higgs voice models through an API and charges by usage, with no flat subscription. Higgs TTS is $0.015 per 1K characters, and generated audio is free, so only the characters you send count (spaces and punctuation included). Higgs Avatar starts at $0.07 per minute of output video for 640x480 and 480x640, and $0.09/min for 640x640; audio and text input are free.
Higgs Realtime, the speech-to-speech model, is estimated at $0.014 per minute of audio output plus about $0.0023/min of audio input, though the real bill comes from tokens consumed. Transcription is $0.0025 per minute of input, with input tokens at $0.75/1M and output at $4.50/1M. New accounts get $10 free credit. Prices checked on 2026-09-29.
| Plan | Price | Billing | Who it is for | Key limits | API |
|---|---|---|---|---|---|
| Higgs TTS API |
$0.02 per 1K characters | Usage annual commitment |
Narration and voiceover | Generated audio free; Spaces and punctuation count as characters | Included |
| Higgs Avatar API |
$0.07 per minute of output video (from) | Usage annual commitment |
Talking avatars | 640x480 and 480x640 at $0.07/min; 640x640 at $0.09/min | Included |
| Higgs Realtime API |
$0.01 per minute audio output (estimate) | Usage annual commitment |
Voice agents | Audio input about $0.0023/min; Actual cost from tokens consumed | Included |
| Transcription API |
$0 per minute of input | Usage annual commitment |
Transcribing audio | Input tokens $0.75/1M, output $4.50/1M | Included |
| If you | Pick | Why |
|---|---|---|
| you are testing the API for the first time | Free credit ($10 for new accounts) | There is no flat plan, so the $10 free credit for new accounts is the way to try Higgs before paying. |
| you need narration or voiceover in many languages | Higgs TTS, $0.015 per 1K characters | Voice cloning from short samples and 100+ languages; you pay only for input characters, not generated audio. |
| you are building a talking avatar | Higgs Avatar, from $0.07/min | Billed per minute of output video; 640x640 costs $0.09/min instead of $0.07/min. |
| you are building a live voice agent | Higgs Realtime, about $0.014/min output | Speech-to-speech with interruption handling; cached input is $0.25/1M tokens, and the final cost follows tokens consumed. |
| you only need speech recognition | Transcription, $0.0025/min | Billed per minute of input audio, with token rates of $0.75/1M input and $4.50/1M output. |
No flat plan; pay per use.
$10 free credit for new accounts.
Estimates exclude applicable taxes; realtime cost billed by tokens.
Excluded from estimates.
Prices checked Sep 29, 2026 on the official US pricing page. Official pricing ↗ Prices and limits change often.
No flat plan; pay per use.
$10 free credit for new accounts.
$0.02 per 1K characters source (opens in a new tab)
Usage, annual commitment
For: Narration and voiceover
Limits
Includes
API included
$0.07 per minute of output video (from) source (opens in a new tab)
Usage, annual commitment
For: Talking avatars
Limits
Includes
API included
$0.01 per minute audio output (estimate) source (opens in a new tab)
Usage, annual commitment
For: Voice agents
Limits
Includes
API included
$0 per minute of input source (opens in a new tab)
Usage, annual commitment
For: Transcribing audio
Limits
Includes
API included
Estimates exclude applicable taxes; realtime cost billed by tokens.
Excluded from estimates.
By usage: TTS $0.015 per 1K characters, avatar from $0.07 per minute, and realtime billed per audio minute or tokens. source (opens in a new tab)
Yes, new accounts receive $10 in free trial credit per the pricing page. source (opens in a new tab)
No, generated audio is free; only input characters are billed. source (opens in a new tab)
No. Our verified data shows pay-per-use pricing only, with no flat plan.
$0.015 per 1K characters. Generated audio is free, and spaces and punctuation count as characters.
New accounts receive $10 free credit.
The estimate is about $0.014 per minute of audio output and about $0.0023/min for audio input, but actual cost comes from tokens consumed. Cached input is $0.25/1M tokens.
No. The estimates exclude applicable taxes.
compare>