Murf AI

Murf's List Voices API endpoint now accepts a model parameter. Pass falcon-2.1 for streaming voices or gen2 for synthesis voices. Each model has its own catalog, so fetch the matching list before sending a voiceId to Stream Speech, WebSockets or Synthesize Speech.

Gen2 has been decommissioned for streaming, but its voices remain available through Synthesize (Gen 2). Use Falcon 2.1 for streaming. Both catalogs are in the Voice Library.

Original source

Murf's Speech to Text is in beta, with real-time transcription over WebSocket. It's available only in the India region and supports 12 languages, including English (India), Hindi and Tamil, plus automatic language detection.

You get partial transcripts while someone speaks and a final transcript with word-level timestamps when you end the turn. Silence doesn't end a turn: your client must send Finalize or CloseStream. Finalize keeps the connection open for the next turn.…

Murf's Falcon 2.1 speech model is live on Stream Speech and WebSockets. You can set exact pronunciations with inline IPA, a phonetic alphabet, or a pronunciation dictionary for words that recur.

Send model: "falcon-2.1" to use it. Existing falcon-2, falcon2 and FALCON identifiers now use Falcon 2.1 automatically.

Murf reports more natural voices, fewer artifacts and more consistent loudness. Pricing is 1 cent per 1000 characters, with a concurrency limit of 5 in US-East and 2 in all other…

Murf's Falcon 2 now lets you control pronunciation with inline IPA or a pronunciation dictionary on Stream Speech and WebSockets.

For an inline pronunciation, put IPA between forward slashes inside double quotes, as in Welcome to "/mɜːrf/".

For every occurrence of a word, use a dictionary entry with its IPA or replacement words. On Stream Speech, pass pronunciationDictionary. On WebSockets, send pronunciationdictionary in voiceconfig once per context.

Original source

Murf AI’s Falcon 2 streaming now supports volumeOffset on Stream Speech and in WebSocket voiceconfig.

Set an integer from -50 to 50 to adjust loudness relative to the model’s default output. The default is 0. Murf says the adjustment adds no latency and lets voice agents match phone lines, earpieces and noisy environments.

Original source

Murf added word-level timestamps to its WebSocket streaming API for Falcon 2.

Set addwordtimestamps to true at the top level of the frame, next to text. Murf says to keep it out of voiceconfig, where it's ignored without an error. Once it's on, it applies to the whole connection. Falcon then sends each spoken word with starts and ends in seconds. These arrive as separate JSON frames mixed in with the audio frames, not inside the base64 audio field.…

Murf AI’s Falcon 2 WebSocket API can now return timing for each spoken word. Set addwordtimestamps to true to enable it for the whole connection. Words arrive in separate JSON frames alongside the audio, with starts and ends values in seconds.

For voice agents handling interruptions, compare word timestamps with the current playback position to work out how much of a response played. The timestamps arrive before the corresponding audio plays, so receiving them doesn’t mean the user has heard…

Murf AI’s Falcon 2 WebSocket API can now return timing for each spoken word. Set addwordtimestamps to true to enable it for the whole connection. Words arrive in separate JSON frames alongside the audio, with starts and ends values in seconds.

For voice agents handling interruptions, compare word timestamps with the current playback position to work out how much of a response played. The timestamps arrive before the corresponding audio plays, so receiving them doesn’t mean the user has heard…

Murf has decommissioned Gen2 on its Streaming API. If you're still using it for streaming speech generation, Murf directs you to migrate to Falcon 2.

Gen2 remains available for non-streaming text-to-speech through the Synthesize Speech endpoint.

Original source