Bulk generation runs a queue of prompts in parallel instead of one image at a time: up to 20 on Solo, 50 on Creator and 300 on Agency. Pick a persona or product, a model and the output settings, check the price on Start, and walk away. Every result lands in Review, where you keep the best and bin the rest.
Up to 20 on Solo, 50 on Creator and 300 on Agency. The limit counts every image still to make across the workspace, so teammates' runs share it. Start takes as many waiting prompts as fit and leaves the rest queued for the next batch.
A new WaveSpeed account runs 2 tasks at a time. Its first top-up, of any size, lifts that to at least 300, which is what lets a batch use your plan's full number.
56 image models that work from reference photos: Muse Image, Flux 2 Klein 9B, Flux 2 Dev, Flux 2 Turbo, Seedream v4, Seedream v5.0 Flash, Kling Image O3, Wan 2.7, MiniMax H3, Nano-Banana, GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, Seedream v4.5, Nano-Banana 2.1, Seedream v5.0 Pro, Nano-Banana 2 Fast, Flux 2 Pro, Nano-Banana 2, GPT Image 2, Flux 2 Max, Grok Imagine v2, Wan 2.7 Pro, Nano-Banana Pro, Nano-Banana Pro Ultra, Flux 3, Microsoft MAI Image 2.5, Microsoft MAI Image 2.5 Pro, Luma Uni v1, Kling Image V3, Qwen Image 3.0, Qwen Image 3.0 Pro, Seedream v5.0 Lite, Nano-Banana 2 Lite, HiDream O1 Image, HiDream O1 Image Dev, Sourceful Riverflow 2.0 Pro, Zhipu GLM Image, Phota, FireRed Image v1.1, JoyAI Image, Meituan LongCat Image, Boogu Image, Flux 2 Flash, Flux 2 Flex, Flux 2 Klein 4B, Vidu Q2, Bria FIBO 1.5, Pruna P-Image, NVIDIA Chrono Edit, Wan 2.6, Wan 2.5, Ideogram Character, Instant Character, Flux PuLID, UNO and USO. For a shot with no reference, 64 text-to-image models write the image from the prompt alone.
Choose a model lists the ones that can take your shot, cheapest first, priced per image for that shot, and tells you how many images to drop for the ones that cannot.
Yes. A Singles shot starts with a persona or a product, and you can add more subjects, a garment from your clothing library and a background. The app tells the model which reference is which: keep this face and body, keep this product's shape, logo and text, dress the person in this exact garment, place them in this setting.
A persona and a product in the same shot take two slots, so they need Creator or above: Solo holds 1 persona or product. Clothing and backgrounds work on every plan.
The Start button shows the run's total before you press it, and each image shows its price for the model, resolution and number of references. Prices are the provider's list rate, lowered when WaveSpeed runs a sale, and the bill goes to your own WaveSpeed account. Personifex adds nothing per image.
Failed images retry automatically in later passes, four by default, and an extra pass costs nothing unless something fails.
Write prompts as cards, paste them raw, or import a text or Markdown file, and duplicates are skipped. Any prompt can be improved or turned into variations with AI, and the Prompt maker writes a whole batch from one idea. Each persona has its own queue, shared with the team.
Prompts run in parallel, not one at a time, up to your plan's concurrency: 20 at once on Solo, 50 on Creator, 300 on Agency, once your WaveSpeed account is past its starter tier. Progress streams live and failed prompts retry automatically in later passes. Wall-clock depends on the model you pick and WaveSpeed's own queue.
Yes. Generation runs on your own WaveSpeed key at their price, and we never mark it up. New WaveSpeed accounts run 2 tasks at a time; your first top-up, of any size, lifts that to at least 300, so a batch can use your plan's full 20, 50 or 300.
Yes. No reference is a Singles option where the prompt describes everything, on 64 text-to-image models. Carousel and the video modes always start from a persona or a product.
It retries in the next pass, up to four passes by default. If it still fails, the prompt shows under Failed in the Prompts history with the reason, ready to queue again.