Use /stream for up to 1,000 characters, time-sensitive cases (e.g. chatbot).
NEW: Use a websocket connection to wss://api.v8.unrealspeech.com/streamWithTimestamps to stream both audio and per-word timestamps.
Use /speech for up to 3,000 characters, which returns publicly accessible URLs to the audio and timestamps.
Use /synthesisTasks for up to 500,000 characters, which returns a task ID you can use to retrieve publicly accessible URLs to the audio and timestamps.
Change the endpoint from https://api.v6.unrealspeech.com to https://api.v8.unrealspeech.com
Update the VoiceID
Everything else is backward-compatible.
\nBelow parameters are used for /stream, /speech and /synthesisTasks endpoints.
VoiceId
American Female: Autumn, Melody, Hannah, Emily, Ivy, Kaitlyn, Luna, Willow, Lauren, Sierra
\nAmerican Male: Noah, Jasper, Caleb, Ronan, Ethan, Daniel, Zane
\nChinese Female: Mei, Lian, Ting, Jing
\nChinese Male: Wei, Jian, Hao, Sheng
\nSpanish Female: Lucía
\nSpanish Male: Mateo, Javier
\nFrench Female: Élodie
\nHindi Female: Ananya, Priya
\nHindi Male: Arjun, Rohan
\nItalian Female: Giulia
\nItalian Male: Luca
\nPortuguese Female: Camila
\nPortuguese Male: Thiago, Rafael
\nBitrate: Defaults to 192k. Use lower values for low bandwidth or to reduce the transferred file size. Use higher values for higher fidelity.
Speed: Defaults to 0. Examples:
0.5: makes the audio 50% faster. (i.e., 60-second audio becomes 42 seconds)
-0.5: makes the audio 50% slower. (i.e., 60-second audio becomes 90 seconds.)
Pitch: Defaults to 1. However, on the landing page, we default male voices to 0.92 as people tend to prefer lower/deeper male voices.
Send up to 1000 characters and a playback stream of audio back in 0.3 seconds. The is the endpoint to use for the fastest response.
\nIn addition to the common parameters, the stream endpoint takes below paramter(s).
Text: This is the text to be synthesized to audio. Up to 1000 characters.
Codec: Defaults to libmp3lame (MP3). Use pcm_mulaw for phone calls. pcm_s16le returns 22050 Hz raw audio.Temperature: Defaults to 0.25. The lower values make audio deterministic and more stable. The higher values make audio more expressive and less-deterministic. With a high Temperature value, audio will be different every time. However, it also increases the probability of mispronunciation.
Send up to 3,000 characters and synchronously receive an MP3 and JSON timestamp URLs.
\nIn addition to the common parameters, the /speech endpoint takes below paramter(s).
Text: This is the text to be synthesized to audio.
TimestampType: By default, the endpoint returns per-sentence timestamps. Use word to get per-word timestamps. The timestamp feature is currently not supported via the /stream endpoint.
Send up to 500,000 characters and immediately receive a TaskId. GET TaskId to check the status. Optimized for cost savings and longer requests. It generally processes ~700 character per seconds. A 500,000-char request takes ~15 minutes and results in ~10 hours of audio.
In addition to the common parameters, the /synthesisTasks endpoint takes below paramter(s).
Text: This is the text to be synthesized to audio.
On average, 800 characters result in 1 minute of audio. In other words, 500,000 characters will result in approximately 10 hours of audio.
\nOn average, it takes 1 second per 800 characters. In other words, 500,000 characters will take approximately 10 minutes.
\nIf you POST a string, it'll return a single OutputUri. If you POST a list of strings, OutputUri will be a list of MP3 URLs.
For longer texts, we recommend sending a list of strings. This way, instead of waiting for the entire task to be completed to access audio, you can start accessing the first audios in the list. i.e., -.mp3
TimestampType: By default, the endpoint returns per-sentence timestamps. Use word to get per-word timestamps. The timestamp feature is currently not supported via the /stream endpoint.
CallbackUrl: If provided, the server will POST a JSON body to the CallbackUrl. A sample body looks like below:
{\n \"TaskId\": \"8282b92d\",\n \"TaskStatus\": \"completed\", // or \"failed\"\n }\n\nCheck the status of a speech synthesis task.
\n","urlObject":{"protocol":"https","path":["synthesisTasks","TaskId"],"host":["api","v8","unrealspeech","com"],"query":[],"variable":[]}},"response":[{"id":"29ebf6c8-e0e7-443c-bed5-d60b7fdcec8d","name":"Completed Request","originalRequest":{"method":"GET","header":[{"key":"Authorization","value":"Bearer API_KEY","type":"text"}],"url":"https://api.v6.unrealspeech.com/synthesisTasks/785010b1"},"status":"OK","code":200,"_postman_previewlanguage":"json","header":[{"key":"Server","value":"nginx/1.18.0 (Ubuntu)"},{"key":"Date","value":"Fri, 01 Sep 2023 22:06:09 GMT"},{"key":"Content-Type","value":"application/json"},{"key":"Content-Length","value":"236"},{"key":"Connection","value":"keep-alive"},{"key":"Access-Control-Allow-Origin","value":"*"}],"cookie":[],"responseTime":null,"body":"{\n \"SynthesisTask\": {\n \"CreationTime\": \"2025-02-13T12:26:01.22Z\",\n \"OutputUri\": [\n \"https://unreal-expire-in-90-days.s3-us-west-2.amazonaws.com/b7cf8b52-c269-4906-8177-22c8fb1010c7-0.mp3\"\n ],\n \"RequestCharacters\": 4,\n \"StatusDetails\": \"N/A\",\n \"TaskId\": \"b7cf8b52-c269-4906-8177-22c8fb1010c7\",\n \"ProcessingTime\": \"0.6402294635772705\",\n \"TaskStatus\": \"completed\",\n \"TimestampsUri\": [\n \"https://unreal-expire-in-90-days.s3-us-west-2.amazonaws.com/b7cf8b52-c269-4906-8177-22c8fb1010c7-0.json\"\n ],\n \"VoiceId\": \"Sierra\"\n }\n}"}],"_postman_id":"95b3b7f4-708b-45e1-8c0e-6421764beea8"}]}