Use a voice service you already have

Connect ElevenLabs, Hume, Microsoft Azure or OpenAI with your organisation's own key, choose a voice, and see what each read will use before it is made.

Voice-overs come from Google, on your organisation's own Google Cloud. If your organisation already pays for another voice service, you can connect it and read films with it instead: ElevenLabs, Hume, Microsoft Azure or OpenAI.

Every read is your organisation's own spend with that service, so Showreel Studio never uses it unless a kit asks for it, and it tells you what a read will use before it is made.

Connect it

Only owners and admins can do this.

1. Make an API key with the service. A restricted key is fine: it needs only to read voices and to make speech. For Microsoft Azure, use your Speech resource's key and note its region (for example uksouth).
2. In the web app, open Organisation and find the Voice services card.
3. Paste the key next to the service (and, for Azure, the region) and press Connect.

The key is checked with the service before it is kept. It is sealed (encrypted) on the service and never shown again; the card shows its last four characters, who added it and when. Disconnect takes it away.

In the terminal:

showreel org voices set-key elevenlabs

showreel org voices set-key azure --region uksouth

This asks for the key without showing it, so it stays out of your shell history.

Choose a voice

List a service's voices, or search its library (ElevenLabs and Hume have one):

showreel voices --service elevenlabs

showreel voices --service elevenlabs --library --accent british --gender female

Hear a voice

In the web app, open Organisation, then Voices… on the Voice services card. Search the voices, press ▶ to hear one, and click its id to copy it for the kit. In the local app, the film's voice panel has ▶ Hear it next to the voice.

ElevenLabs voices come with a free preview. Hume, Microsoft Azure and OpenAI have none, so hearing one reads a short sentence (about 80 characters) on your account; you are asked first, and hearing it again the same day costs nothing.

In the terminal:

showreel voices --service elevenlabs --preview VOICE_ID

Hear your own film's lines

Hear a few lines of your film in several voices before you choose (three lines each, unless you pick them with --lines):

showreel voice my-film --audition VOICE_ID_1,VOICE_ID_2

Read a film with it

In the kit's voice, set the service and the voice:

"voice": { "service": "elevenlabs", "voice": "VOICE_ID", "settings": { "speed": 0.95 } }

Use "service": "hume", "azure" or "openai" the same way. Azure's voices are names like en-GB-RyanNeural; OpenAI's are marin, cedar and the rest.

Then voice the film as usual. Each service reads the script as one performance, so the level and tone hold from the first line to the last:

• ElevenLabs reads it in one go, and the lines are cut where it says each one is spoken.
• Hume reads it in one go and hands each line back on its own. Your film's voice direction becomes Hume's acting description.
• Microsoft Azure is sent the script with an exact pause between lines, and the read is split on them. A speaking style (such as chat) stands in for a direction.
• OpenAI is asked to stop between lines, and the read is split on the stops. Your film's voice direction becomes its instructions.

What a read uses

Before reading, Showreel Studio says how many characters the read uses. A read above your organisation's limit asks first; at the terminal answer y, or add --yes. Owners and admins set the limit on the Voice services card. A take already read with the same words and settings is kept and costs nothing to use again.

If a word comes out wrong

Add it to the kit's pronounce list, written as it should be said:

"pronounce": { "Bros.": "Brothers" }

This changes only what the voice reads, never the captions.

What next

• Choose a voice
• Voice takes