AI Rate CardHow to work out the cost of AI images, video and voice: a calculator with live prices
Updated Nov 7, 2026
Image, video and voice models are rarely priced per token: they are sold per image, per second of video, per million characters of speech or per minute of audio. The images, video and voice calculator turns your job into money at every model that sells it by its own unit, the cheapest offer of each model, in your currency.
How to use the calculator
- Pick the job: images, video, text to speech or speech to text.
- Enter how much: images, seconds of video, characters to speak, or minutes of audio.
- Read the table: one row per model - its cheapest seller, the price per unit and the cost of your job, cheapest first. A row marked "under review" has a price that moved or does not match the seller's page; check it on the card.
Examples from today's prices
- 100 images: from $0.05 (FLUX.1 schnell at DeepInfra); the typical model costs about $3. 92 models are priced per image.
- 60 seconds of video: from about $0.15; the typical model costs about $6.
- 100,000 characters of speech: from about $0.10 (Kokoro 82M); the typical model about $2.
- 10 hours of transcription (600 minutes): from about $0.12 (Whisper Large v3 Turbo, Qwen3 ASR and Nemotron ASR at DeepInfra); the typical model about $1.30.
Before you choose the cheapest row
- Resolution and length: a seller may charge more for a larger image or longer clip - the conditions are on the offer card.
- Quality: the cheapest image model is a fast draft model; compare a few rows on your own prompts.
- Token-priced models: some image models (GPT Image, Gemini) bill per token and are not in this table - price them with the cost calculator.