{"observation":{"id":"862ea299-ef77-4932-baf5-89190b469aae","tool":"leonardo-ai","tool_name":"Leonardo AI","criterion":"identity-preservation","criterion_name":"Identity preservation","criterion_definition":"How closely the generated face matches the reference image across scenes.","criterion_evidence_type":"transformation","criterion_rank_role":"decisive","criterion_rank_role_reason":"This is the core of the task: the tool must keep the same character looking like the reference image across scenes. (3 of 3 judges)","scenario":"full-frontal-portrait","scenario_name":"Full frontal portrait","group_tag":null,"scenario_description":"Full frontal portrait reference image with fair skin, curly dark hair, bindi, gold jhumka earrings, and a green stone necklace. All features are clearly visible in good natural lighting, making it the easiest identity anchor for the tools.","modality":"image","input_text":null,"input_artifact_refs":[],"stresses":["Baseline identity preservation","Accessory retention","Best-case frontal face matching","Consistent character reuse across varied scenes"],"verdict":"failed","score":null,"score_total":null,"note":"It can render the desert action scene cleanly while replacing the reference with a different character whose face, proportions and overall appearance no longer match.","evidence_state":"verified","source":null,"artifacts":[{"url":"https://d3epheqghktydj.cloudfront.net/leonardo-ai-input-1-5e817fd0f4a6.png","role":"input","alt":null},{"url":"https://d3epheqghktydj.cloudfront.net/leonardo-ai-leonardo-input1-horseride-f769fa286b59.jpg","role":"output","alt":null}],"run_id":"1dfb8fa4-f7f0-47dc-911b-7db3af467e9a","study_title":"Generate Consistent AI Characters Across Different Scenes and Poses","study_kind":"generation","research_task":"86b96df11","tested_at":null,"completeness":"input-and-output","input":{"state":"not-captured","text":null,"files":[],"modality":"image","stresses":["Baseline identity preservation","Accessory retention","Best-case frontal face matching","Consistent character reuse across varied scenes"]},"tool_page_slug":"leonardo-ai","tool_url":"https://aidemos.com/tools/leonardo-ai","permalink":"https://aidemos.com/evidence/862ea299-ef77-4932-baf5-89190b469aae","api_url":"https://ai.aidemos.com/v1/observations/862ea299-ef77-4932-baf5-89190b469aae"},"peers":[{"id":"163f2da9-8a7d-4d76-bc0b-a4ed35e52693","tool":"chatgpt","tool_name":"ChatGPT","verdict":"struggled","score":null,"score_total":null,"note":"Identity drops to only a 50–70% match in an action scene, with significant facial-structure drift away from the reference even though the body styling remains plausible.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/chatgpt-chatgpt-input1-horseride-dbbcd9f8517e.png","evidence_url":"https://aidemos.com/evidence/163f2da9-8a7d-4d76-bc0b-a4ed35e52693"},{"id":"af4776d3-f013-4967-a4a8-bd0db677c8e4","tool":"gemini","tool_name":"Gemini","verdict":"failed","score":null,"score_total":null,"note":"In the horse-riding scene, the tool can erase the reference face entirely: the report calls the identity match very weak and says the output is a completely different character with changed face shape, eyes, and eyebrows.","artifact_count":2,"thumbnail":"https://d3epheqghktydj.cloudfront.net/gemini-input-1-abf743cbaaf0.png","evidence_url":"https://aidemos.com/evidence/af4776d3-f013-4967-a4a8-bd0db677c8e4"},{"id":"669f1e69-857a-4f41-88cc-29961c0ee861","tool":"imagineart","tool_name":"ImagineArt","verdict":"worked","score":null,"score_total":null,"note":"The interrogation-room output is the strongest Input 1 match, with the eyes, nose, lips, and face shape closest to the reference across the Input 1 scenes.","artifact_count":2,"thumbnail":"https://d3epheqghktydj.cloudfront.net/imagineart-input-1-561e7b836cdd.png","evidence_url":"https://aidemos.com/evidence/669f1e69-857a-4f41-88cc-29961c0ee861"},{"id":"ed66c473-a880-4d06-8d75-0990ec076849","tool":"scenario","tool_name":"Scenario","verdict":"worked","score":null,"score_total":null,"note":"The tool can keep a subject recognisably the same person in an action scene, with face shape, eyes, and overall look staying close to the reference.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/scenario-scenario-input1-horseride-d01d0d9cfea4.png","evidence_url":"https://aidemos.com/evidence/ed66c473-a880-4d06-8d75-0990ec076849"}],"other_criteria":[{"id":"2f9fb0e6-bf73-43da-9569-51dac9414373","criterion":"accessory-and-detail-retention","criterion_name":"Accessory & detail retention","rank_role":"decisive","verdict":"struggled","score":null,"score_total":null,"note":"It drops the reference's natural curl pattern and smooths hair and skin detail into a more stylised finish.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/2f9fb0e6-bf73-43da-9569-51dac9414373"},{"id":"b57aa471-7b05-4433-a4f2-f1a621c5aaf3","criterion":"expression-accuracy","criterion_name":"Expression accuracy","rank_role":"decisive","verdict":"failed","score":null,"score_total":null,"note":"It does not translate an explicitly angry or guarded prompt into facial expression, defaulting to calm neutrality.","artifact_count":1,"evidence_url":"https://aidemos.com/evidence/b57aa471-7b05-4433-a4f2-f1a621c5aaf3"},{"id":"504bb238-1d54-4cff-bc33-9bd9b9bb9b59","criterion":"scene-compliance","criterion_name":"Scene compliance","rank_role":"decisive","verdict":"mixed","score":null,"score_total":null,"note":"It can place the subject in a believable interrogation setup with the correct outfit, but the room stays too neat and loses the harsh institutional atmosphere.","artifact_count":1,"evidence_url":"https://aidemos.com/evidence/504bb238-1d54-4cff-bc33-9bd9b9bb9b59"}],"appears_in":[{"page_type":"ranking","slug":"consistent-ai-characters","title":"Best AI Tools for Consistent AI Characters Across Scenes and Poses","url":"https://aidemos.com/best/consistent-ai-characters","binding":"run"}],"same_scenario":[{"id":"163f2da9-8a7d-4d76-bc0b-a4ed35e52693","tool":"chatgpt","tool_name":"ChatGPT","verdict":"struggled","score":null,"score_total":null,"note":"Identity drops to only a 50–70% match in an action scene, with significant facial-structure drift away from the reference even though the body styling remains plausible."},{"id":"af4776d3-f013-4967-a4a8-bd0db677c8e4","tool":"gemini","tool_name":"Gemini","verdict":"failed","score":null,"score_total":null,"note":"In the horse-riding scene, the tool can erase the reference face entirely: the report calls the identity match very weak and says the output is a completely different character with changed face shape, eyes, and eyebrows."},{"id":"669f1e69-857a-4f41-88cc-29961c0ee861","tool":"imagineart","tool_name":"ImagineArt","verdict":"worked","score":null,"score_total":null,"note":"The interrogation-room output is the strongest Input 1 match, with the eyes, nose, lips, and face shape closest to the reference across the Input 1 scenes."},{"id":"ed66c473-a880-4d06-8d75-0990ec076849","tool":"scenario","tool_name":"Scenario","verdict":"worked","score":null,"score_total":null,"note":"The tool can keep a subject recognisably the same person in an action scene, with face shape, eyes, and overall look staying close to the reference."}]}