Skip to main content

ElevenLabs Voice Consistency

ACT3 AI uses ElevenLabs as the primary voice generation engine for dialogue, narration, and character speech. ElevenLabs provides the voice consistency that professional productions require — the same character sounds the same across every line of dialogue in every scene, regardless of tone, emotion, or phrasing.

Why Voice Consistency Matters​

In a traditional production, a voice actor records all their lines in one or a few sessions, and the voice stays consistent throughout the film. In AI-generated productions, dialogue can be generated line by line across many sessions. Without a consistent voice engine, the same character can sound noticeably different from scene to scene.

ElevenLabs solves this by assigning each character a Voice Profile — a stable voice identity that generates consistent output every time that character speaks.

How It Works in ACT3 AI​

  1. Each digital actor is assigned a voice profile in the Voice tab of the Actor configuration
  2. All dialogue for that actor is sent to ElevenLabs using that voice profile
  3. ElevenLabs generates the audio and ACT3 AI syncs it to the actor's lip movement in the generated video
  4. The result is a character whose voice sounds identical whether they are speaking in Act 1 or Act 3

Assigning an ElevenLabs Voice Profile​

  1. Open the Actor Library and select an actor
  2. Click the Voice tab
  3. Browse the ElevenLabs Voice Library — hundreds of professional voice profiles categorized by:
    • Gender expression
    • Age range
    • Accent and language
    • Character type (authoritative, warm, mysterious, energetic, etc.)
  4. Click Preview to hear a sample
  5. Click Assign to lock that voice to the actor

The assigned voice generates all dialogue for that actor going forward, with consistent character and tone.

Emotion and Delivery Control​

ElevenLabs supports emotional delivery tags in dialogue. In the script or shot dialogue field, you can add emotion markers:

  • [nervous] — the line is delivered with a nervous affect
  • [angry] — forceful, raised energy
  • [whispered] — quiet, breathy delivery
  • [sad] — slower, heavier tone
  • [excited] — higher energy, faster pace

ACT3 AI passes these markers to ElevenLabs automatically when generating dialogue audio. The markers appear as optional hints in the Script editor as you write.

Multi-Language Support​

ElevenLabs supports over 30 languages. To set a character's dialogue language:

  1. In the actor's Voice tab, select Language
  2. Choose from the available ElevenLabs-supported languages
  3. All generated dialogue for that actor renders in the selected language

For multi-language productions, each language version of a character can use the same base voice profile with the language switched — giving the same voice character across different dubbed versions.

ACT3 AI Internal Voice Engine​

In addition to ElevenLabs, ACT3 AI offers an Internal Voice Engine — a proprietary system built for fast, lower-cost dialogue generation.

The internal engine is optimized for:

  • Speed — faster generation for draft and preview passes
  • Cost — lower credit cost per line compared to ElevenLabs
  • Volume — high-throughput dialogue generation for projects with hundreds of dialogue lines

The internal engine is well-suited for:

  • Draft dialogue passes before final production
  • Projects with very high dialogue volume where budget is a factor
  • Background characters and extras with minimal speaking lines

For full internal engine documentation, see Internal Voice Engine.

When to Use Each Engine​

ScenarioRecommended Engine
Lead character dialogue — final productionElevenLabs
Supporting character dialogue — key scenesElevenLabs
Background charactersInternal Engine
Draft dialogue for timing and reviewInternal Engine
Final export with maximum qualityElevenLabs
High-volume content (100+ dialogue lines)Internal Engine (draft), ElevenLabs (final)

ElevenLabs Voice Cloning (Custom Voices)​

For productions that require a custom voice — matching a specific performer's voice — ACT3 AI supports ElevenLabs Voice Cloning:

  1. Submit a recording request to a voice actor via the Voice Casting workflow
  2. The voice actor records a reference sample (5–10 minutes of clean audio)
  3. Upload the sample to ACT3 AI under Voice Library → Clone Voice
  4. ElevenLabs creates a custom voice model from the recording
  5. The cloned voice is available as a private voice profile in your Organization

Custom voice cloning requires the voice actor's explicit consent and appropriate rights agreement.

Credit Usage​

ElevenLabs dialogue generation consumes credits separately from video rendering:

TypeCost
Standard voice (per line)0.5 credits
Premium ElevenLabs voice (per line)1 credit
Internal engine (per line)0.1 credits
Custom cloned voice (per line)1.5 credits

Dialogue credits are deducted when audio is generated. Re-generating the same line (with edits or retakes) consumes credits again.

Troubleshooting​

Voice sounds inconsistent between scenes — Confirm the same voice profile is assigned to the actor across all scenes. Check the Voice tab in the Actor Library.

Emotion tags not applied — Verify the tag format matches the expected syntax ([emotion]) and is placed at the start of the dialogue line.

Voice not available for a language — Not all ElevenLabs voices support all languages. Check language availability in the Voice Library filter.

Audio out of sync with lip movement — Re-sync by regenerating the shot or manually adjusting the audio offset in the editor's Audio panel.