Does the face match the voice?

MiguelCompare Miguel’s likeness with his references. A realistic face alone is not a match.

One portrait and voice. Two workflows.

Test details

2 raw attempts · 2 models

Fixed: Approved Saturn portrait and ElevenLabs voice excerpt.
Changed: Sync-3 and MiniMax H3 Max Lip Sync workflows.
Check: Compare face drift and mouth timing separately.

Two released image-and-audio workflows, without visual text prompts. Different workflows; fresh AI reviews are pending.

Recorded October 7, 2026. References are shown per attempt.

Sort

How it sorts

Within each test: highest complete AI rating, then lower historical cost. Partial panels stay unranked by rating. Lowest cost reverses that priority. Unknown values follow known ones. Ranges use their upper estimate. Human votes stay separate.

Video uses USD per delivered minute. Images use USD per generation until complete spend and human usability checks establish cost per usable image. Failed submissions and retries stay in the test history. Fees and taxes are unconfirmed.

MiniMax H3 Max Lip Sync ↗ · 2

References used

Reference details

Approved starting image for the video workflow.

Est. $4.80/minEstimated · $0.8 for this test

Cost basis

Historical quote for the requested 10-second workflow; actual billed cost, fees and taxes unconfirmed. Sync tier range retained where applicable; not a current quote.

AI scoreReview neededRecords found

Previous AI reviews

0 of 5 judges returned usable numeric scores. The combined score waits for five. Unknown results are not zero. AI diagnostics; community votes and Miguel’s verdict stay separate.

Archived source records found. Prompt and reference associations need review before a corrected judging packet is qualified. The earlier panel used an incomplete packet (5/5 reviews received). Fresh reviews have not run yet. This workflow used an image and recorded speech, without a visual text prompt.

ClaudeUnsure—
anthropic/claude-sonnet-5.5
The person's identity, shirt and build are well preserved and the video is visually clean. However, the brief is missing, and the setting and framing change substantially in the first 2s, so suitability can't be confirmed.
GeminiUnsure—
google/gemini-4-argon
The generated video maintains good consistency with the reference image in facial identity, body proportions, and visual quality during a smooth zoom-in while the subject speaks. However, because the original brief is missing and audio is excluded, suitability for the intended prompt cannot be confirmed.
GrokNo—
x-ai/grok-4.7
The output is visually polished and identity-stable, but the unsupported floating seat is a major physics defect. The missing brief prevents shot-compliance review, and the abrupt framing change weakens continuity.
QwenYes—
qwen/qwen3-vl-235b-a22b-instruct
All visible dimensions (face, body, visual, physics, temporal) meet target with no major defects. Brief compliance is unknown due to missing brief. Output is stable and coherent across sampled frames, with consistent identity and environment. No artifacts or motion errors observed.
MistralUnsure—
mistralai/mistral-large-4-0
The generated video shows strong visual fidelity to the single reference image: consistent face, body, clothing, and environment across 10 seconds. Natural expression changes (speaking, smiling, blinking) are present. However, the original brief is unavailable, so shot accuracy and instruction compliance cannot be assessed. Minor AI artifacts in hands and occasional facial softness prevent a full

Video frame reviews cover sampled visuals; they do not certify audio, lip sync or every frame.

Earlier reviews · 1

Gemini 3.8 Flash · Unknown score
Earlier visual rubric · 2026-10-13. Earlier model/rubric; excluded from this five-judge score.

CommunityVotes are recorded separately below

Voting help

If your browser blocks embedded sign-in, open this review separately ↗. Votes and Credits use the same Navo account.

Prompt, settings & test details

Same approved portrait scene and voice excerpt; different workflow.

Preliminary operator note: camera zoom may violate locked framing; human scoring pending.

MiniMax H3 Max Lip Sync · fal.ai

Actual billed cost unconfirmed; do not equate quote with charge.

Sync-3 ↗ · 1

References used

Reference details

Approved starting image for the video workflow.

Est. $6.42–$7.98/minEstimated · $1.07 – $1.33 for this test

Cost basis

Historical quote for the requested 10-second workflow; actual billed cost, fees and taxes unconfirmed. Sync tier range retained where applicable; not a current quote.

AI scoreReview neededRecords found

Previous AI reviews

0 of 5 judges returned usable numeric scores. The combined score waits for five. Unknown results are not zero. AI diagnostics; community votes and Miguel’s verdict stay separate.

ClaudeUnsure—
anthropic/claude-sonnet-5.5
The output is a stable, identity-consistent talking-head clip of a seated man on a space-ring background. Quality is good. The brief is missing, so fit for its purpose cannot be confirmed.
GeminiUnsure—
google/gemini-4-argon
The output preserves the reference subject's likeness, proportions, and environment well with coherent speaking motions, though the missing brief prevents a final suitability decision.
GrokNo—
x-ai/grok-4.7
The clip is visually stable, but the unsupported seated pose is a major physics defect. Missing brief and reference evidence also prevent likeness and shot-accuracy approval.
QwenUnsure—
qwen/qwen3-vl-235b-a22b-instruct
All visible dimensions (face, body, visual, physics, temporal) meet their targets with no observable defects. However, brief rating is null due to missing original prompt, making shot accuracy and instruction compliance unverifiable. Recommendation is unsure because core task alignment cannot be confirmed without the brief.
MistralUnsure—
mistralai/mistral-large-4-0
Generated video shows strong visual similarity to the reference image with preserved identity, clothing, and environment. Face and body match well with minor AI artifacts. However, brief is null so instruction compliance cannot be assessed. Physics rating is mixed due to implausible floating scenario without visible support. Temporal continuity is stable. Without the original brief, cannot confirm

Video frame reviews cover sampled visuals; they do not certify audio, lip sync or every frame.

Earlier reviews · 1

Gemini 3.8 Flash · Unknown score
Earlier visual rubric · 2026-10-13. Earlier model/rubric; excluded from this five-judge score.

CommunityVotes are recorded separately below

Voting help

If your browser blocks embedded sign-in, open this review separately ↗. Votes and Credits use the same Navo account.

Prompt, settings & test details

Same approved portrait scene and voice excerpt; different workflow.

Preliminary operator note: mouth openings may be exaggerated; human scoring pending.

Sync-3 · direct Sync.so

Actual billed cost unconfirmed; do not equate quote with charge.

Results

Two image-and-audio workflows, with no visual text prompt. Compare face stability and lip timing; fresh AI reviews are pending.

Method, costs & limits

Exploratory tests with unchanged outputs and recorded failures. Model names are visible before voting; votes do not establish a model-wide winner.

Check face drift and lip timing separately. Two released clips are a small screening set, not evidence about every shot or voice.

Controlled human acceptance and a general recommendation remain unresolved.

More tests →

Try this shot

Storyboard a talking-head shot using your own character and voice. Approve costs before rendering.

Does the face match the voice?

References used

References used

Results

Try this shot

AGV Report