The tool can miss the requested emotion entirely, producing a neutral and emotionless face instead of the prompted angry and guarded look.
What was measured
Expression accuracy
Whether the emotional tone and facial expression match what was explicitly prompted.
decisive for this rankingtransformation
If the character’s intended emotion or facial expression does not match the prompt, the generated character is not being controlled reliably across scenes. (3 of 3 judges)
What was given, what came back
Test input: Three-quarter face portrait · image
Input — what we sent
Input 2.png
Input not captured
This run recorded no prompt or input file for the test, so we cannot show you what produced the result below. Capture gaps are tracked, not hidden.
Three-quarter face reference image with medium-dark skin, tight curly hair in an updo, bindi, and a floral dress. The partial angle and softer lighting make it a harder reference than the frontal portrait and are meant to stress identity consistency.
Why this input is hard
- · Identity preservation with partial face angle
- · Hair texture and updo retention
- · Skin tone fidelity
- · Reference-image difficulty under softer lighting
Output — unretouched
Also checked on this input — same tool, 5 other criteria
Accessory & detail retention✓ WorkedA busy market scene can still preserve fine clothing and hair details, keeping the sari, blouse, hairstyle, and curly hair texture recognizable.Identity preservation✓ WorkedIn the crowded market scene, the character stays highly recognisable, preserving face shape, smile, eyebrows, and overall facial structure.Identity preservation✓ WorkedWith a 3/4 reference, the interrogation-room output still preserves identity well, keeping face shape, skin tone, curly hair texture, eyebrows, and overall structure close to the input.Scene compliance✓ WorkedThe crowded market prompt is followed strongly, with a busy outdoor background, realistic crowd density, market stalls, and a natural walking pose with tote bag.Scene compliance✓ WorkedThe interrogation-room prompt is followed well, including the table, plain wall, overhead light, formal clothing, and clinical lighting.
Provenance
- Observation
- 7d21adab-f4e6-4f0a-9975-37fabe9dc1c2
- Evidence run
- 1dfb8fa4-f7f0-47dc-911b-7db3af467e9a
- Study
- Generate Consistent AI Characters Across Different Scenes and Poses
- Research task
- 86b96df11
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "gemini"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 3 other tools
measured on Expression accuracy
ChatGPT✓ WorkedDelivers the requested cold, guarded, unsmiling expression with the exact vibe the prompt asked for.ImagineArt✓ WorkedThe interrogation-room output preserves the reference expression well.Leonardo AI✗ FailedThe same neutral-expression failure repeats on a second reference, so the tool still misses the prompt's angry, guarded tone.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com