Models / Canopy Labs
Orpheus 3B
by Canopy Labs
Reads your text aloud.
Orpheus 3B turns written text into spoken audio. It is made by Canopy Labs. On Tīrtha it reads up to 300 characters per call.
Price per 1M characters
The live price, read from billing as this page loads. You pay for the characters of text you send, from prepaid credit. There is no charge for the audio it returns. Requests that fail are not charged.
Live price not available just now.
tirtha/orpheus-3b-0.1-ftGood at
- Natural deliveryThe maker built it for natural intonation, emotion and rhythm.
- Emotion tagsSimple tags in your text guide emotion and intonation.
- Built on LlamaA fine-tune of Llama 3.2 3B, so the Llama 3.2 licence and use policy apply.
Model card details
- Made by
- Canopy Labs
- Kind
- Text to speech
- Licence
- Apache-2.0. Built with Llama. Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved. Use follows the Llama 3.2 Acceptable Use Policy (licences).
- Listed on Hugging Face
- Mar 17, 2025
- Languages
- English
- Parameters
- 3 billion
- Base
- Llama 3.2 3B
- Preset voices
- 7
Hear it
Made with Orpheus 3B. Real outputs, not edited, each with the exact text it read and the voice it used.
Welcome back. Your order shipped this morning and should arrive on Friday. Reply to this message if you need anything else.AI-generatedBuilt with Llama
I found the missing puzzle piece under the couch this morning.AI-generatedBuilt with Llama
If the river stays this calm, we can paddle out to the island.AI-generatedBuilt with Llama
Voices
Other voices the model offers include these, by language. We have not tested them.
- English)
- Tara
tara, Leahleah, Jessjess, Leoleo, Dandan, Miamia, Zaczac
Open weights, Apache-2.0 licence. Maker facts from Canopy Labs's model card on Hugging Face (huggingface.co/canopylabs/orpheus-3b-0.1-ft). Limits are the ones Tīrtha enforces today.
Request and response
Fields you send
| Field | Required | Values |
|---|---|---|
model | yes | tirtha/orpheus-3b-0.1-ft |
input | yes | The text to speak, 1 to 300 characters. |
voice | no | One of the voice ids listed under Voices. Default tara. |
response_format | no | wav only, which is the default. |
Fields not in this table are refused with unknown_parameter. That includes speed, which no model here takes.
What comes back
HTTP/1.1 200 OK
Content-Type: audio/wav
Cache-Control: no-store
<the wav audio file, as bytes>The body is the audio file itself, ready to save or play. There is no JSON wrapper and no link to fetch. You are billed for the characters of input you sent.
If something goes wrong
Errors come back as JSON: {"error": {"code", "message", "type"}}. A call that fails is not charged.
| HTTP | Code | What it means |
|---|---|---|
| 400 | invalid_input | The text is empty or over the limit: split it. |
| 400 | invalid_voice | The voice is not one this model has. |
| 400 | invalid_response_format | Only wav is served. |
| 400 | unknown_parameter | A field the API does not take. Send only the fields above. |
| 400 | invalid_request | The body is not a JSON object. |
| 400 | content_policy | Refused by content policy. Nothing was charged. |
| 400 | request_rejected | The request was rejected. Nothing was charged. |
| 402 | billing_required | This needs a prepaid credit account. Add credits in your dashboard first. |
| 402 | insufficient_credits | Not enough prepaid credit for this request. The message says how much it may need. |
| 404 | model_not_found | The model id is wrong, or is not this kind of model. |
| 404 | model_not_priced | This model is not available yet. |
| 429 | media_busy | This model is busy right now. Retry in a few seconds. Nothing was charged. |
| 503 | billing_unavailable | Billing is temporarily unavailable. Please retry. |
| 503 | unavailable | Temporarily unavailable. Nothing was charged. |
What a job costs
An article of about 1,000 words (6,000 characters) costs about -
Worked out from the live price above, at about 6 characters a word, counting spaces. Punctuation counts too.
How to use it
Pick it when expressive English speech from a named voice matters more than speed.
Model id tirtha/orpheus-3b-0.1-ft · POST https://api.tirtha.ai/v1/audio/speech
- One call reads up to 300 characters. For longer text, split it at the end of a sentence, make one call per part, and play or join the files in order. The Python example does the splitting, and cuts a single over-long sentence at a space.
- Pick a voice with
voice: tara, leah, jess, leo, dan, mia, zac. - The audio comes back as WAV, the only format served, so
response_formatcan be left out. WAV keeps full quality; convert it on your side if you need a smaller file. - The response body is the audio file. Save it straight to disk with the right extension, or stream it to a player.
- Check how the voice reads numbers, dates and short forms such as 10:30 or Dr. Write them out the way they should sound if it matters.
- Tell your listeners the voice is AI-generated, for example in your app or at the start of the audio.
Set TIRTHA_API_KEY to a key from your dashboard first. Keys, credits and the base URL: Quickstart.