UGC vs influencer is the file you buy
UGC vs influencer is the file you buy: a file plus rights, not a following. Spark and whitelist are a separate contract.
8 min read
Claude Opus 5.5 Release Date, Price and Benchmarks
Claude Opus 5.5 was released on 22 September 2026: $4 in, $20 out per million tokens, Fable 5.1-level results, about 40% cheaper than Opus 5.
8 min read
Export the Sora library before delete
Export at sora.chatgpt.com/sunset before OpenAI deletes Sora-associated data. Remaining ChatGPT/Sora credits go to Codex, not video.
7 min read
Flux 3 T2V is native audio to 20s
Flux 3 T2V is 34 credits for 5s, up to 136 at 20s, native audio. Black Forest Labs GA on 4 Aug 2026. One row, one job.
7 min read
Omni Video generates; Flash Edit edits
Omni Video generates at 9 credits (T2V/I2V/R2V, 4K, 4–10s). Flash Edit is 13 credits, edit-video, audio false. We do not publish Gemini Omni 1.1 Flash.
6 min read
GPT-6 Sol and Luna: what launched and what it costs
GPT-6 Sol and Luna launched on 22 September 2026: Sol costs $2 in and $10 out per million tokens, Luna $0.10 and $0.50. GPT-6.1 Sol followed 29 Sept.
7 min read
Jev is a System One model, not an LLM
Jev is a System One model from TypeSafe AI, not an LLM. It returns a typed value and a calibrated probability from program state. It does not generate text.
5 min read
Jev: zero hallucinations means out of schema
Jev's zero hallucinations means zero out-of-schema outputs, not zero wrong answers. TypeSafe's own benchmark is ~68%, and the references are frontier models.
5 min read
Kling O3 Pro I2V vs T2V
Both Kling O3 Pro rows are 70 credits, native audio, 3–15s. I2V is the still as contract. T2V invents the frame.
6 min read
Kling V3 4K T2V vs O3 Pro
Kling V3 4K T2V is 168 credits; O3 Pro is 56. Most Reels do not need 4K. Pick O3 Pro unless the master is 4K.
8 min read
LTX 2.3 Fast: from 20 credits, 4K, audio
LTX 2.3 Fast starts at 20 credits, native audio, up to 4K, 6–20s. Use it for cheap 4K sound takes. Skip it when SKU must hold or the hero is dialogue.
7 min read
Qwen 4: what Alibaba announced at Apsara 2026
Qwen 4 was announced at Apsara on 22 September 2026 but is not released: no weights, price or date as of 8 October. What is confirmed, and what to use now.
8 min read
Qwen-Image-2.1 generates a real alpha channel
Qwen-Image-2.1, released 20 September 2026, generates native RGBA with a true alpha channel up to 2K from a prompt, replacing generate-then-background-remove.
5 min read
Qwen-Image-2.1 is not Apache 2.0
Qwen-Image-2.1 is not Apache 2.0. The Qwen Research License bars commercial use without a separate agreement. The $0.03-$0.075 API is a different model.
5 min read
Qwen-Image-2.1 runs on one consumer GPU
Qwen-Image-2.1 cuts the visual generator from 20B to 7B. BF16 weights are 14.2 GB, INT8 is 7.26 GB, and a Q8 run sits near 16 GiB, up to 17 GB at 50 steps.
6 min read
Replace Sora drafts with Kling Turbo
Cheap silent drafts go to Kling 3 Turbo: 22 credits, 3–15s, 1080p, audio false. Then TTS or a native-audio hero. Turbo has no native audio.
7 min read
Replace Sora SKU shots with Seedance
SKU identity that has to hold maps to Seedance 2.0 at 69 credits, 4–15s, 1080p native audio, or Fast R2V at 25 credits, 720p.
7 min read
Seedance 2.0 Fast R2V is 55 credits
Seedance 2.0 Fast R2V is 55 credits (was 78), 720p, native audio, 4–15s. Full Seedance 2.0 starts at 44. Fast is the draft/ref row.
6 min read
Seedance 2.5 720p vs 2.0 1080p
Seedance 2.5 is 47 credits, 4–30s, 720p native audio. Seedance 2.0 is 69 credits, 4–15s, 1080p. Pick duration or resolution.
6 min read
Map Sora 2 durations to Versely rows
Sora 2 T2V is 4/8/12/16/20s. Map those lengths to Veo 4/6/8, Seedance 4–15 or 4–30, Flux 3 5–20, Grok Imagine 6–30.
8 min read
Sora 2 Pro storyboard after the API
The 10/15/25s Sora 2 Pro I2V storyboard row dies with the API. Replace by duration and stills: Seedance, Flux 3, Veo first-last, movie-mode.
7 min read
Sora API last-day checklist, 24 Sept
On 24 Sept 2026 the Sora API still answers until OpenAI pulls it. Export masters, stop new prompts, reroute remaining jobs.
8 min read
Ten references in one Qwen-Image-2.1 edit
Qwen-Image-2.1 composes up to 10 reference images into one scene by prefix KV cache reuse. Long prompts overload the encoder and add unsolicited Chinese text.
5 min read
Typed decisions for platform policy checks
An Instagram AI-label or unoriginal-Reel check is a fixed schema. Jev returns a typed value and a confidence score from text. It cannot see the video.
6 min read