Callirra

@Callirra

Callirra
@Callirra
Sep 12
The prompt for this clip asked for wind over open ground, hooves on dry gravel, creaking tack and a distant bird call. MiniMax H3 rendered all of it in the same pass as the picture. No separate sound step. Ten measured clips ↓
00:00
1
Callirra
@Callirra
Sep 12
MiniMax H3 renders the picture and the audio in the same pass. No separate sound step, no parameter to turn it off. This clip: 768x1344, 24fps, 6.6s — the AAC track came back with it. 10 renders, with settings, wall time and cost ↓
00:00
1
Callirra
@Callirra
Sep 11
GPT Image 2.5 is really two models: Flare (photographic) and Sunburst (stylised). Same prompt limit, same 16 reference images — but two separate price ladders. The cheapest tier starts at 2 credits. Which to use, and when: medium.com/@callirra/gpt-…
Callirra
@Callirra
Sep 11
We priced 10 seconds of AI video across the current model catalogue. Cheapest route: 30 credits. Seedance 2.5 at 720p: ~388 credits. Same duration. 13x spread. The full per-second breakdown: medium.com/@callirra/what…
Callirra
@Callirra
Sep 11
Every frame below came from one prompt of 371–553 words — canvas, light, materials, exact text, object counts, all pinned down before generating. The same repo also ships 5-word prompts. 50 prompts for GPT Image 2.5, each with the frame it produced. Free, CC BY 4.0. Repo ↓
1