Seedance 1.0 Pro Fast: the short answer
fal documents separate text-to-video and image-to-video endpoints for Seedance 1.0 Pro Fast. That confirms an access route through fal, not availability through every provider or through Token Forge Cloud. ByteDance’s Seedance 1.0 research describes the broader model but does not establish Pro Fast’s variant-specific performance.
For a concrete cost comparison, fal lists an approximate $0.245 per 5-second, 1080p text-to-video generation for Pro Fast and about $0.62 for Seedance 1.0 Pro. That is about 60% less at those listed settings—but it is a listed-price comparison, not proof of equal output quality or lower latency for your workload.
Practical decision: consider Pro Fast for cost-sensitive iteration if its documented endpoint fits your workflow. Before production, pilot matched prompts and settings, measuring quality, end-to-end latency, failures and retries, and total billed cost. Do not treat the general Seedance 1.0 research timing as a Pro Fast API benchmark.
What the model name does—and does not—verify
ByteDance’s Seedance 1.0 materials describe the broader model and its research claims; they do not establish that every general Seedance 1.0 capability or performance result applies to the Pro Fast API variant. The paper, for example, reports 41.4 seconds to generate a 5-second 1080p video on an NVIDIA L20. That is a research-system result—not a Pro Fast API latency measurement or service-level guarantee.
A third-party CometAPI article describes Pro Fast as oriented toward lower latency and compute cost, but provides no comparative measurements or pricing in the captured text. Treat that as unmeasured positioning, not a verified speed or savings guarantee.
Check task fit and output requirements
fal documents distinct text-to-video and image-to-video endpoints. In its image-to-video schema, prompt and image_url are required inputs. Listed aspect ratios are 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or auto. The schema lists resolution choices of 480p (faster generation), 720p (balance), and 1080p (higher quality). The captured schema excerpt identifies duration as an optional enum but does not provide its allowed values. Confirm the current duration options before building around a particular clip length.
The paper describes multi-shot storytelling for Seedance 1.0 and gives examples such as shot-reverse-shot dialogue and cut-in/cut-away sequences. These are research-paper examples—not tested Pro Fast outputs or guarantees of character and shot continuity in a particular provider deployment.
Production API access and Fast-versus-Pro costs
What fal documents for API access
fal lists these endpoint IDs:
- Text-to-video:
fal-ai/bytedance/seedance/v1/pro/fast/text-to-video - Image-to-video:
fal-ai/bytedance/seedance/v1/pro/fast/image-to-video
For image-to-video, fal’s example shows a POST request to the endpoint, an Authorization: Key $FAL_KEY header, and a JSON body with prompt and image_url. The reference presents a raw cURL request separately from a JavaScript SDK subscribe example. Values such as result.data and result.requestId belong to that SDK example; the reference does not establish them as fields in the raw HTTP response. A request ID tracks a job—it is not itself the generated video. The available excerpt does not establish the exact generated-video artifact field, so check the current result schema before wiring up storage or downstream processing.
fal’s general queue documentation describes submitting a request and using the returned request ID and status/result URLs to track it, then retrieving the result after completion. That generic queue flow is distinct from the model-specific output schema. The available sources do not establish Pro Fast-specific usage limits or all failure and retry conditions.
The available sources do not establish that Token Forge Cloud exposes this exact Pro Fast model identifier. Confirm the model ID, endpoint, supported inputs, output schema, queue or polling behavior, usage limits, and price with the provider you intend to use rather than inferring them from fal’s documentation.
Listed price comparison—not a speed or quality benchmark
fal lists approximately $0.245 per 5-second 1080p text-to-video generation for Pro Fast and $0.62 per 5-second 1080p generation for Seedance 1.0 Pro. At those listed rates, 1,000 generations would be about $245 versus $620, or roughly $375 less before retries or other charges. This is an illustration based on fal’s listed prices, not a promise of your invoice total.
That comparison supports a potential listed-price advantage on fal at the stated settings. It does not establish matched quality, faster completion, or total production economics. Include retries, failed jobs, different settings, and any provider-specific charges in your pilot estimate.
For context only, the Seedance 1.0 paper reports 41.4 seconds to generate a 5-second 1080p video on an NVIDIA L20. That result is not identified as Pro Fast and is not an API response-time measurement.
Who should choose it?
- Consider Pro Fast if fal’s documented endpoint fits your workflow and its listed price matters. Pilot your own prompts before sending substantial volume.
- Compare it with standard Pro if a higher-priced option might improve the results you care about. Keep prompts and settings consistent, then compare quality and end-to-end time.
- Pause before production if you require Token Forge Cloud specifically or depend on a duration value or output-artifact field not confirmed in the endpoint schema. Verify those details with your provider first.
The selection answer is conditional: fal documents Pro Fast API endpoints and lists a lower 5-second 1080p price than standard Pro, but the available evidence does not establish a general speed advantage, equivalent quality, or Token Forge Cloud availability. Choose it after checking endpoint fit and piloting your workload—not on the word “Fast” alone.
Sources
Bytedance Seedance V1 Pro Fast API — fal Documentation
Seedance 1.0 Pro Fast Image-to-Video Schema — fal
Seedance 1.0 Pro Fast Text-to-Video API — fal
Seedance 1.0 Pro Text-to-Video API — fal
Seedance 1.0: Exploring the Boundaries of Video Generation Models