Keep a character consistent. Choose a model. Understand the cost. Practical image and video guides to help you make your next piece.
Given a single reference photo, can a model put the same person in a new scene? We ran GPT Image 2, Nano Banana 2, and Seedream 5.0 Lite under identical conditions and recorded latency, cost, and identity retention.
Given a single reference photo, can a model put the same person in a new scene? We ran GPT Image 2, Nano Banana 2, and Seedream 5.0 Lite under identical conditions and recorded latency, cost, and identity retention.
Measured per-image cost for GPT Image 2, Nano Banana 2, and Seedream 5.0 Lite — plus the quality setting that takes one model from $0.019 to $0.1299 per image.
A measured review of OpenAI's GPT Image 2 — latency, cost, the parameters it accepts, and the instructions it handles badly. Only what we confirmed by running it.
A measured review of Google's Nano Banana 2 (Gemini 3.1 Flash Image) — the speed, how literally it follows angle instructions, and the one-image-per-run limit.
A measured review of ByteDance Seed's Seedream 5.0 Lite — the flat $0.035 per image, the resolution floor that returns HTTP 400, and its failure to hold a referenced person.
We measured billing for Wan 3.0 and Seedance 2.5. Two seconds at 480p lists at $0.10 and billed at $0.2125. Why video estimates have to come from measurement.
Do you need to train a LoRA to keep the same person or character? We compare attaching a reference against training, on cost, setup, and what happens when you switch models.
Why an attached reference still comes back as someone else. The three causes are model choice, which image you picked, and the sentences you should have left out.
How to pin a character's look in text when you cannot attach a reference image — which attributes to make explicit, and which sentences to leave out.
A front-side-back turnaround needs two things at once: the angle you asked for, and three views staying aligned. Measured, those are the strengths of two different models.
From choosing a reference to changing angles and outfits. Start with the identity comparison, then choose a guide for your workflow.
GUIDE 02Compare reference fidelity, speed and instruction following alongside cost. Shortlist with the comparison, then read each model’s strengths and limits.
GUIDE 03Understand image and video costs and prepare references to reduce rework. Prices in reports are snapshots of the measurement date; check current terms before deciding.
