Eve library

ISIS v2 Voice Line Pack Consistency-Locked Workflow Graph

Guide

Revision 1

On this page

Generated: 2026-03-03T12:14:27.462Z Checklist: MXA-0545 (ISIS2-00743) Asset Family: voice-line-pack Phase: Phase 4 | Milestone: M3 Launch (2026-07-27 to 2026-10-30)

Objective#

Define a consistency-locked voice-line-pack workflow graph that enforces continuity profile state and voice consistency profile state across speaker identity, pronunciation, prosody, and ambience dimensions while retaining deterministic synthesis defaults.

Continuity Profile#

Profile ID Minimum Passing Dimensions Locked Dimensions
scene-continuity-profile 4 environment, lighting, lens, color-script, motion

Deterministic Defaults#

Setting Value
modelName xtts_v2_voiceset_a.safetensors
promptTemplateId voice-line-pack-consistency-locked-v1
styleProfileId voice-line-pack-continuity-anchor-v1
seed 2026073105
durationSeconds 8
speakingRate 1
pitchShiftSemitones 0
voiceStability 0.76
voiceConsistencyStrength 0.78
sampleRateHz 48000
targetLufs -18
outputFormat wav
outputChannels 1
outputBitDepth 24
continuityProfileId scene-continuity-profile
continuityProfileRevision 1
voiceConsistencyProfileId VOICE_LINE_CONTINUITY_PROFILE_V1
voiceConsistencyProfileRevision 1
voiceConsistencyProviderMode required
consistencyWindowSize 5
continuityLockStrength 0.82
voiceLockStrength 0.8
prosodyLockStrength 0.76
pronunciationLockStrength 0.72
ambienceLockStrength 0.68
phonemeAlignmentTargetScore 0.9
intelligibilityTargetScore 0.92
integratedLufsDeltaTolerance 1
prosodyDriftTolerance 0.12

Input Contract#

Input Type Required Default
continuity_profile_id string yes "scene-continuity-profile"
continuity_profile_revision integer yes 1
continuity_locked_dimensions string[] yes ["environment","lighting","lens","color-script","motion"]
continuity_notes string yes "Lock speaker identity, prosody drift, ambience bed, and pronunciation coherence against approved continuity profile."
continuity_window_size integer yes 5
voice_consistency_profile_id string yes "VOICE_LINE_CONTINUITY_PROFILE_V1"
voice_consistency_profile_revision integer yes 1
voice_consistency_provider_mode enum yes "required"
voice_id string yes "narrator_f"
speaker_embedding_id string no "speaker_embed_narrator_f_v1"
speaker_embedding_url string no "asset://references/embeddings/speaker_embed_narrator_f_v1.json"
speaker_embedding_vector array no [0.15,-0.03,0.29,0.08]
speaker_reference_audio_url string yes "asset://references/audio/voice-line-speaker-lock-001.wav"
prosody_reference_audio_url string no "asset://references/audio/voice-line-prosody-lock-002.wav"
ambience_reference_audio_url string no "asset://references/audio/voice-line-ambience-lock-003.wav"
pronunciation_reference_audio_url string no "asset://references/audio/voice-line-pronunciation-lock-004.wav"
speaker_reference_audio_urls array no ["asset://references/audio/voice-line-speaker-lock-001.wav","asset://references/audio/voice-line-prosody-lock-002.wav"]
use_speaker_reference boolean yes true
use_prosody_reference boolean yes true
use_ambience_reference boolean yes true
use_pronunciation_reference boolean yes true
continuity_lock_strength float yes 0.82
voice_lock_strength float yes 0.8
prosody_lock_strength float yes 0.76
pronunciation_lock_strength float yes 0.72
ambience_lock_strength float yes 0.68
phoneme_alignment_target_score float yes 0.9
intelligibility_target_score float yes 0.92
integrated_lufs_delta_tolerance float yes 1
prosody_drift_tolerance float yes 0.12
cue_id string yes "VOICE_LINE_M3_CONSISTENCY_LOCKED_01"
script_text string yes "Commander, reroute auxiliary power to deck three and hold for extraction."
speaker_id enum yes "narrator_f"
language_locale enum yes "en-US"
emotion enum yes "confident"
speaking_rate float yes 1
pitch_shift_semitones float yes 0
duration_seconds integer yes 8
pronunciation_lexicon_ids array no ["LEXICON_EN_US_OPERATIONS_V1"]
pronunciation_dictionary_locators array no ["asset://references/lexicons/en-us-operations-v1.pls"]
style_tokens array no ["cinematic","clear-diction","controlled-cadence"]
voice_consistency_strength float yes 0.78
voice_stability float yes 0.76
text_prompt_hint string no "Preserve speaker identity and pronunciation clarity while maintaining continuity-locked cadence."
prompt_template_id string yes "voice-line-pack-consistency-locked-v1"
style_profile_id string yes "voice-line-pack-continuity-anchor-v1"
sample_rate_hz integer yes 48000
target_lufs float yes -18
output_format enum yes "wav"
model_name string yes "xtts_v2_voiceset_a.safetensors"
seed integer yes 2026073105

Graph Nodes#

Node ID Class Role Output Count
n1 AudioConsistencyProfileLoader continuity-profile-load 1
n2 VoiceConsistencyProfileLoader voice-consistency-profile-load 1
n3 LoadAudio speaker-reference-load 1
n4 LoadAudio prosody-reference-load 1
n5 LoadAudio ambience-reference-load 1
n6 LoadAudio pronunciation-reference-load 1
n7 VoiceLinePromptComposer prompt-construction 1
n8 VoiceLineModelLoader model-load 1
n9 VoiceProfileLockMerge continuity-lock-merge 2
n10 VoiceLineSynthesisNode audio-generation 1
n11 VoiceLineCleanupDenoise audio-post 1
n12 AudioLoudnessNormalize audio-post 1
n13 MetadataManifestBuilder metadata-build 1
n14 SaveAudio artifact-output 0
n15 SaveJSON manifest-output 0
n16 SaveArchive archive-output 0

Graph Edges#

Edge ID From To
e1 n1:0 n9.continuity_profile
e2 n2:0 n9.voice_consistency_profile
e3 n3:0 n9.speaker_reference
e4 n4:0 n9.prosody_reference
e5 n5:0 n9.ambience_reference
e6 n6:0 n9.pronunciation_reference
e7 n7:0 n9.prompt
e8 n9:0 n10.conditioned_prompt
e9 n9:1 n10.continuity_controls
e10 n8:0 n10.model
e11 n10:0 n11.audio
e12 n11:0 n12.audio
e13 n12:0 n14.audio
e14 n1:0 n13.continuity_profile
e15 n2:0 n13.voice_consistency_profile
e16 n10:0 n13.sampling_trace
e17 n13:0 n15.payload
e18 n12:0 n16.primary_artifact
e19 n13:0 n16.metadata_manifest

Topological Node Order#

n1 -> n2 -> n3 -> n4 -> n5 -> n6 -> n7 -> n8 -> n9 -> n10 -> n11 -> n12 -> n13 -> n14 -> n15 -> n16

Research Alignment#

Workflow Path Control Nodes Audio Nodes Profile Nodes Reference Audio Inputs Node Classes
Voice Performance Synthesis libs/isis/workflows/src/workflows/audio/voice-performance-synthesis.json 0 3 0 yes AudioDeesser, KokoroModelLoader, KokoroPromptBuilder, KokoroTextToSpeech, SaveAudio
SFX Sound Designer libs/isis/workflows/src/workflows/audio/sfx-sound-designer.json 0 5 0 yes AudioLoudnessNormalize, SaveAudio, StableAudioNode, StableAudioOpenModelLoader, StableAudioPromptComposer
audio-sfx-pack-consistency-locked docs/domains/isis/remediation-v2/ISIS_V2_AUDIO_SFX_PACK_CONSISTENCY_LOCKED_WORKFLOW_GRAPH.json 0 11 2 no AudioConsistencyProfileLoader, AudioLoudnessNormalize, AudioProfileLockMerge, LoadAudio, MetadataManifestBuilder, SaveArchive, SaveAudio, SaveJSON, StableAudioNode, StableAudioOpenModelLoader, StableAudioPromptComposer
social-cutdown-pack-consistency-locked docs/domains/isis/remediation-v2/ISIS_V2_SOCIAL_CUTDOWN_PACK_CONSISTENCY_LOCKED_WORKFLOW_GRAPH.json 4 0 3 no CLIPTextEncode, CheckpointLoaderSimple, ConsistencyProfileLoader, ControlNetApply, ControlNetLoader, EmptyLatentImage, IPAdapterApply, IPAdapterFaceID, IPAdapterLoader, KSampler, LoadImage, MetadataManifestBuilder, ProfileLockMerge, SaveImage, SaveJSON, VAEDecode

Acceptance Criteria#

  • Graph is acyclic and topologically sortable.
  • Continuity profile support is explicit through AudioConsistencyProfileLoader and VoiceProfileLockMerge nodes.
  • Voice consistency profile support is explicit through VoiceConsistencyProfileLoader and voice consistency contract inputs.
  • Speaker/prosody/pronunciation continuity lock paths are represented with deterministic lock strengths and optional fallbacks.
  • Quality policies include SM-005 (phoneme/intelligibility) and SM-006 (loudness/prosody drift) threshold targets.
  • Metadata manifest and project archive outputs are explicitly represented in output contract and graph topology.
  • Output contract aligns to MXA-0541 required output classes for voice-line-pack.

Traceability#

  • Checklist item: MXA-0545
  • Ticket key: ISIS2-00743
  • Source references:
    • docs/domains/isis/remediation-v2/ISIS_REMEDIATION_V2_TICKETS.json
    • docs/domains/isis/remediation-v2/ISIS_V2_VOICE_LINE_PACK_REQUIREMENTS_BRIEF.json
    • docs/domains/isis/remediation-v2/ISIS_V2_VOICE_LINE_PACK_BASELINE_WORKFLOW_GRAPH.json
    • docs/domains/isis/remediation-v2/ISIS_V2_VOICE_LINE_PACK_IMAGE_CONDITIONED_WORKFLOW_GRAPH.json
    • docs/domains/isis/remediation-v2/ISIS_V2_VOICE_LINE_PACK_CONTROL_CONDITIONED_WORKFLOW_GRAPH.json
    • docs/domains/isis/remediation-v2/ISIS_V2_CONSISTENCY_TAXONOMY.json
    • libs/isis/workflows/src/workflows/audio/voice-performance-synthesis.json
    • libs/isis/workflows/src/workflows/audio/sfx-sound-designer.json
    • docs/domains/isis/remediation-v2/ISIS_V2_AUDIO_SFX_PACK_CONSISTENCY_LOCKED_WORKFLOW_GRAPH.json
    • docs/domains/isis/remediation-v2/ISIS_V2_SOCIAL_CUTDOWN_PACK_CONSISTENCY_LOCKED_WORKFLOW_GRAPH.json
    • docs/domains/isis/remediation-v2/ISIS_V2_AUDIO_CONDITIONING_SUPPORT.md
    • docs/domains/isis/remediation-v2/ISIS_V2_VOICE_CONSISTENCY_PROFILES.md
    • docs/domains/isis/remediation-v2/ISIS_V2_OUTPUT_PACKAGE_TAXONOMY.json
    • ISIS_TODOS_2.md