Record a voiceover from the chat
By Steven Van ·
The Voice bar offers 30 Gemini voices, six starting presets and a Compare 3 option, with WAV or MP3 downloads.
Bearly now has a Voice bar under the latest message in a chat. It lets people record a voiceover, podcast intro or narration from the conversation, and pick a voice by hearing it read their own script. Bearly described the change in its text-to-speech documentation.
Listening to an answer
Selecting Hear this reply on the Voice bar, or Read aloud beneath an answer, plays the answer aloud. Bearly records it in the current voice and keeps the take in the chat, so it can be played or downloaded later. Typing /tts in a chat with some text and a note on how it should sound adds an audio player to the conversation.
Recording a voiceover
Opening the Voice bar starts with the question "What are you making?", with six answers: Video voiceover, Podcast intro, Audiobook, Proofread by ear, Announcement and Wind-down. Each sets a model, a lead voice, two alternates and a written delivery direction, and all of it can be edited. Video voiceover, for example, starts with an informative narrator and a clear, measured pace, while Audiobook starts with a warm storyteller and gentle pauses.
- Script: type or paste text, or select Use reply for Bearly's latest answer or Use my message for what was sent.
- Voice: each card in the Cast list has a play button that reads the first sentence of the script with the written delivery. Voices suggested for the chosen job come first, marked with a dot.
- Delivery: describe how it should sound, such as "warm and unhurried", or pick a suggestion.
- Generate: one take, or Compare 3 to record the same script in three voices at once. Earlier takes are listed below the player.
As a take plays, the words highlight in turn. The timing is estimated, so it can drift slightly from the voice.
Models, downloads and usage
A switch at the top of Voice sets the model. Fast uses Gemini Flash-Lite and suits drafts and proofreading. Studio uses Gemini Flash for the most expressive delivery. OpenAI uses GPT-4o Mini with its own voices. Fast and Studio share the same 30 voices.
Gemini takes download as WAV files and OpenAI takes as MP3. Turning on Use for Read Aloud makes Read aloud use the chosen model, voice and delivery. Voice samples and takes count toward AI usage, though replaying a sample with the same words and direction does not make another generation. Team administrators can turn text to speech off for members with a team policy.
